Skip to main content

Full text of "Seismic hazard characterization of 69 nuclear plant sites east of the Rocky Mountains"

See other formats


DOC, 

Y 3.N88; 

25/5250/V.6 NUREG/CR-5250 

UCID-21517 
Vol. 6 



Seismic Hazard Characterization 
of 69 Nuclear Plant Sites 
East of the Rocky Mountains 



Regional Comparison Between Sites, Site Effects, 
General Discussion, and Conclusions 



Prepared by D.L. Bernreuter, J.B. Savy, R.W. Mensing, J.C. Chen 
Lawrence Livermore National Laboratory 



Prepared for 

U.S. Nuclear Regulatory 

Commission 



NOTICE 

This report was prepared as an account of work sponsored by an agency of the United States 
Government. Neither the United States Government nor any agency thereof, or any of their 
employees, makes any warranty, expressed or implied, or assumes any legal liability of re- 
sponsibility for any third party's use, or the results of such use, of any information, apparatus, 
product or process disclosed in this report, on'represents that its use by such third party would 
not infringe privately owned rights. 

- UNIVERSITY OF 

I LLINOIS LIBRARY 

AT URBANACHAMPAIGN 
GOOKSTACKG 



NOTICE 
Availability of Reference Materials Cited in NRC Publications 
Most documents cited in NRC publications will be available from one of the following sources: 

1. The NRC Public Document Room, 1717 H Street, N.W. 
Washington, DC 20555 

2. The Superintendent of Documents, U.S. Government Printing Office, Post Office Box 37082, 
Washington, DC 20013-7082 

3. The National Technical Information Service, Springfield, VA 22161 

Although the listing that follows represents the majority of documents cited in NRC publications, 
it is not intended to be exhaustive. 

Referenced documents available for inspection and copying for a fee from the NRC Public Docu- 
ment Room include NRC correspondence and internal NRC memoranda; NRC Office of Inspection 
and Enforcement bulletins, circulars, information notices, inspection and investigation notices; 
Licensee Event Reports; vendor reports and correspondence; Commission papers; and applicant and 
licensee documents and correspondence. 

The following documents in the NUREG series are available for purchase from the GPO Sales 
Program: formal NRC staff and contractor reports, NRC-sponsored conference proceedings, and 
NRC booklets and brochures. Also available are Regulatory Guides, NRC regulations in the Code of 
Federal Regulations, and Nuclear Regulatory Commission Issuances. 

Documents available from the National Technical Information Service include NUREG series 
reports and technical reports prepared by other federal agencies and reports prepared by the Atomic 
Energy Commission, forerunner agency to the Nuclear Regulatory Commission. 

Documents available from public and special technical libraries include all open literature items, 
such as books, journal and periodical articles, and transactions. Federal Register notices, federal and 
state legislation, and congressional reports can usually be obtained from these libraries. 

Documents such as theses, dissertations, foreign reports and translations, and non-NRC conference 
proceedings are available for purchase from the organization sponsoring the publication cited. 

Single copies of NRC draft reports are available free, to the extent of supply, upon written 
request to the Division of Information Support Services, Distribution Section, U.S. Nuclear 
Regulatory Commission, Washington, DC 20555. 

Copies of industry codes and standards used in a substantive manner in the NRC regulatory process 
are maintained at the NRC Library, 7920 Norfolk Avenue, Bethesda, Maryland, and are available 
there for reference use by the public. Codes and standards are usually copyrighted and may be 
purchased from the originating organization or, if they are American National Standards, from the 
American National Standards Institute, 1430 Broadway, New York, NY 10018. 



NUREG/CR-5250 
UCID-21517 
Vol. 6 



Seismic Hazard Characterization 
of 69 Nuclear Plant Sites 
East of the Rocky Mountains 



Regional Comparison Between Sites, Site Effects, 
General Discussion, and Conclusions 



Manuscript Completed: November 1988 
Date Published: January 1989 

Prepared by 

D.L. Bernreuter, J.B. Savy, R.W. Mensing, J.C. Chen 



Lawrence Livermore National Laboratory 
7000 East Avenue 
Livermore, CA 94550 



Prepared for 

Division of Engineering and System Technology 

Office of Nuclear Reactor Regulation 

U.S. Nuclear Regulatory Commission 

Washington, DC 20555 

NRC FIN A0448 



7>DC, 



p::'-:^^y/'^ 



Abstract 

The EDS Seismic Hazard Characterization Project (SHC) is the outgrowth of an 
earlier study performed as part of the U.S. Nuclear Regulatory Commission's 
(NRC) Systematic Evaluation Program (SEP). The objectives of the SHC were: 
(1) to develop a seismic hazard characterization methodology for the region 
east of the Rocky Mountains (EUS), and (2) the application of the methodology 
to 69 site locations, some of them with several local soil conditions. The 
method developed uses expert opinions to obtain the input to the analyses. An 
important aspect of the elicitation of the expert opinion process was the 
holding of two feedback meetings with all the experts in order to finalize the 
methodology and the input data bases. The hazard estimates are reported in 
terms of peak ground acceleration (PGA) and 5% damping velocity response 
spectra (PSV). 

A total of eight volumes make up this report which contains a thorough 
description of the methodology, the expert opinion's elicitation process, the 
input data base as well as a discussion, comparison and summary volume 
(Volume VI). 

Consistent with previous analyses, this study finds that there are large 

uncertainties associated with the estimates of seismic hazard in the EUS, and 

it identifies the ground motion modeling as the prime contributor to those 
uncertainties. 

The data bases and software are made available to the NRC and to public uses 
through the National Energy Software Center (Argonne, Illinois). 



-m- 



Digitized by the Internet Archive 

in 2013 



http://archive.org/details/seismichazardcha06bern 



Table of Contents 
Volume VI 



PAGE 



Abstract iii 

Table of Contents v 

List of Tables and Figures vi 

Foreword xiv 

List of Abbreviations and Symbols xvii 

Executive Summary: Volume VI xxi 

SECTION 1 INTRODUCTION 1 

SECTION 2 DISCUSSION OF THE SENSITIVITY OF THE COMPUTED SEISMIC 12 
HAZARD TO SEVERAL IMPORTANT ASPECTS OF THE METHODOLOGY 
USED TO ESTIMATE THE GROUND MOTION 

2.1 Background 12 

2.2 Correction for the Site's Soil Category 12 

2.3 Sensitivity to G-Expert 5's Model 34 

2.4 Apparent Disconnect Between the PGA Hazard 47 
and the Spectral Hazard 

SECTION 3 COMPARISON BETWEEN SITES AND REGIONAL OBSERVATIONS 51 

3.1 General Comparisons Between Sites 51 

3.2 Regional Comparisons (PGA) 64 

3.3 Regional Spectral Comparisons 82 

SECTION 4 SUMMARY OF RESULTS AND CONCLUSIONS 88 

APPENDIX A References A-1 

APPENDIX B Maps of Seismic Zonation for Each of the 11 S-Experts B-1 



-V- 



List of Tables and Figures 



PAGE 



Table 1.1a Sites and Soil Category Used for Each 3 

Site in Batch 1. 

Table 1.1b Sites and Soil Category Used for Each 4 

Site in Batch 2 

Table 1.1c Sites and Soil Category Used for Each 5 

Site in Batch 3 

Table l.ld Sites and Soil Category Used for Each 6 

Site in Batch 4 

Table 2.2.1 Definition of the Eight Site 16 

Categories 

Table 2.2.2 Nearby Sites in Different Soil 17 

Categories 

Table 3.1.1 Key for Sites in Figure 3.1.2a and 58 

3.1.2b 

Table 3.1.2 Key for Sites in Figure 3.1.2a and 59 

3.1.2b and Median Values at 0.2g 

Table 3.1.3 Key for Sites in Figure 3.1.3 and 60 

Median Hazard Values for 0.6g 

Table 3.2.1 Rock Sites Selected for the Comparison 69 

Between the Hazard for Sites Located 
in New England, near Charleston, near 
New Madrid, and Half-Way Between New 
Madrid and Charleston 

Table 3.2.2 List of Sites with Some Structures Founded 70 

on Rock and Some on Shallow Soil 

Table 3.2.3 Ratios of PGA Values Between Shallow and Rock 72 
Conditions for Fixed Values of the Hazard 



Figure 1.1a Map showing the location of the Batch 
1 sites contained in Vol. II of this 
report. Map symbols are given in 
Table 1.1a. 



-VI- 



PAGE 



Figure 1.1b Map showing the location of the Batch 8 

2 sites contained in Vol. Ill of this 
report. Map symbols are given in 
Table 1.1b. 

Figure 1.1c Map showing the location of the Batch 9 

3 sites contained in Vol. VI of this 
report. Map symbols are given in 
Table 1.1c. 

Figure l.ld Map showing the location of the Batch 10 

4 sites contained in Vol. V of this 
report. Map symbols are given in 
Table l.ld. 

Figure 1.2 Map giving the relative location of 11 

all the sites included in this study. 

Figure 2.2.1 Smoothed median correction factors for 18 

the Till-like categories listed in 
Table 2.2.1 relative to rock. The PGA 
correction factors are plotted at 
0.01s. 

Figure 2.2.2 Smoothed median correction factors for 18 

the Sand-like categories listed in 
Table 2.2.1 relative to the rock. The 
PGA correction factors are plotted at 
0.01s. 

Figure 2.2.3 Simple correction factors selected by 19 

G-Expert 5. 

Figure 2.2.4 Comparison between the CPHCs for the 20 

case when the Limerick site's soil 
category is rock and the case when it 
is considered to be deep soil. 

Figure 2.2.5 Comparison between the AMHCs and the 21 

BEHCs for the case when the Limerick 
site's soil category is rock and the 
case when it is deep soil. 

Figure 2.2.6 Comparison between the CPUHS with a 22 

10,000 year return period for the case 
when the Limerick's soil category is 
rock and the case when it is deep 
soil. 



-vn- 



PAGE 

Figure 2.2.7 Comparison between the CPHCs for the 23 

case when the Limerick site's soil 
category is considered to be: Til 1-1, 
Til 1-2, and Til 1-3. For comparisons 
the rock is also plotted 

Figure 2.2.8 Comparison between the median 10,000 24 

year return period CPUHS for the case 
when the Limerick's site soil category 
is considered to be: Till-1, Till-2 
and Til 1-3. For comparison the rock 
case is also plotted. 

Figure 2.2.9 Same as Fig. 2.2.8 except the 15th and 25 

85th percentile CPUHS are also 
plotted. 

Figure 2.2.10 Comparison between the CPUHS for the 26 

cases when the Limerick site's soil 
category is considered to be: Sand-1, 
Sand-2 and Sand-3. For comparison the 
rock case is also plotted. 

Figure 2.2.11 Comparison between the median 10,000 27 

year return period CPUHS for the cases 
when the Limerick site's soil category 
is considered to be: Sand-1, Sand-2 
and SAND-3. For comparison, the rock 
case is also plotted. 

Figure 2.2.12 Same as Fig. 2.2.11 except the 15th 28 

and 85th percentile CPUHS are also 
plotted. 

Figure 2.2.13a Comparison of the median CPHCs between 29 

the Vermont Yankee site (rock) and the 
nearby Yankee Rowe site {till-2). 

Figure 2.2.13b Comparison of the median CPHCs between 30 

the Braidwood site (rock) and the 
nearby Lasalle site {till-2). 

Figure 2.2.13c Comparison between the median CPHCs 31 

for the Kewaunee site {till-2) and the 
Point Beach site (till-2). 

Figure 2.2.14 Comparison between the CPHCs for the 32 

case when the Browns Ferry site is 
rock and the case when it is treated 
as a deep soil site. 



-vm- 



PAGE 



Figure 2.2.15 Comparison between the AMHCs and BEHCs 33 

for the Browns Ferry site ran as a 
rock site and as a deep soil site. 

Figure 2.3.1a Comparison between the BEHCs per G- 39 

Expert for S-Expert 3's input for the 
Browns Ferry site. 

Figure 2.3.1b Comparison between BEHCs per G-Expert 40 

for S-Expert 4's input for the Browns 
Ferry site. 

Figure 2.3.2a Comparison between the CPHCs when all 41 

5 G-Experts are used and when G- 
Experts are not included for the 
Browns Ferry site. 

Figure 2.3.2b Comparison between the AMHCs and BEHCs 42 

when all 5 G-Experts are used and when 
G-Expert 5 is not included for the 
Browns Ferry site. 

Figure 2.3.3 Comparison between the CPHCs when all 43 

the G-Experts are used and when G- 
Expert 5 is not included for the 
Limerick site. 

Figure 2.3.4 Comparison between CPHCs when all of 44 

the G-Experts are used and when G- 
Expert 5 is not included for the River 
Bend site. 

Figure 2.3.5 Comparison between the CPHCs when all 45 

of the G-Experts are used and when G- 
Expert 5 is not included for the Salem 
site. 

Figure 2.3.6 Comparison between the 10,000 year 46 

return period CPUHS when all of the G- 
Experts are used and when G-Expert 5 
is not included for the Browns Ferry 
site. 

Figure 2.4.1 Comparison between the 10,000 year 48 

return period CPUHS for the Limerick 
site and the spectral value estimated 
from the CPUHS using the Newmark-Hall 
amplification of 1.0 at 0.03 sec. to 
convert acceleration to relative 
velocity. 



-IX- 



PAGE 

Figure 2.4.2 Comparison between the spectra 49 

obtained using the random vibration 
spectral model RV-5RS at an epicentral 
distance of 15km for magnitudes 5, 6 
and 7 and the spectra obtained using 
the Newmark-Hall median spectral 
amplifications applied to the random 
vibration velocity model RV-5V. The 
models RV-5A and 5V are consistent 
with the spectral model RV-5RS. 

Figure 2.4.3 Illustration of how the spectral 50 

values for periods shorter than 0.04s 
can be estimated. The appropriate PGA 
value is read from off the CPHCs at 
the appropriate return period, 
converted to spectral velocity and 
plotted at 0.01s. Then a straight 
line, as shown, is used to connect the 
PGA value to the last computer CPUHS 
value at 0.04s. 

Figure 3.1.1a Comparison of the median CPHCs for all 53 

the sites listed in Tables l.la-d. 

Figure 3.1.1b Comparison of the median CPHCs for the 54 

sites in Vol. II. The plot symbols 
used to identify the sites are given 
in Table 1.1a. 

Figure 3.1.1c Comparison of the median CPHCs for the 55 

sites in Vol. III. The plot symbols 
used to identify the sites are given 
in Table 1.1b. 

Figure 3.1. Id Comparison of the median CPHCs for the 56 

sites in Vol. IV. The plot symbols 
used to identify the sites are given 
in Table 1.1c. 

Figure 3.1.1e Comparison of the median CPHCs for the 57 

sites in Vol. V. The plot symbols 
used to identify the sites are given 
in Table l.ld. 



-X- 



PAGE 



Figure 3.1.2a A plot of the log of the annual 61 

probability of exceeding 0.2g for all 
the sites in Vols. II-V. The plot 
symbols are: M-arithmetic mean, (*) 
for the 15th and 85th percentiles and 
B-best estimate. The sites are 
ordered by Volume. The key is given 
in Table 3.1.1 and cross referenced 
with Table 3.1.2. 

Figure 3.1.2b Same as Fig. 3.1.2a except the sites 62 

have been ordered by median 
probability of exceeding 0.2g. The 
ordering is given in Table 3.1.2 and 
cross referenced in Table 3.1.1. 

Figure 3.1.3 A plot of the log of the annual 63 

probability of exceeding 0.6g for all 
of the sites in Vols. II-V. The sites 
have been ordered by the median 
probability of exceeding 0.6g. The 
plot symbols are: M-median, 
A-arithmetic mean, (*)-15th and 85th 
percentiles and B»best estimate. The 
ordering is given in Table 3.1.3. 

Figure 3.2.1 The location of the sites listed in 73 

Table 3.2.1 is shown by the symbol X 
relative to the historic New Madrid 
(NM) and Charleston (C) earthquakes. 
The 10,000 year return period PGA 
median g-values are also shown. 

Figure 3.2.2 A plot of the log of the annual 74 

probability of exceeding 0.2g for each 
of the sites in Table 3.2.1. The site 
number is given in Table 3.2.1. The 
plot symbols are: M=median, 
A=arithmetic mean, {*)=15th and 85th 
percentiles and B=best estimate. 

Figure 3.2.3 A plot of the estimated 1000 year 75 

return period PGA only including large 
earthquakes of magnitude 6.5 and 
greater based on the BEHCs for each 
site listed in Table 3.2.1. 



-XT- 



PAGE 

Figure 3.2.4a BEHCs which include only the 76 

contribution to the PGA hazard from 
earthquakes within the indicated 
magnitude range for the Arkansas site 
(near the New Madrid area). 

Figure 3.2.4b BEHCs which include only the 77 

contribution to the PGA hazard from 
earthquakes within the indicated 
magnitude range for the Seabrook site, 
in New England (far from either the 
New Madrid or the Charleston areas). 

Figure 3.2.4c BEHCs which include only the 78 

contribution to the PGA hazard from 
earthquakes within the indicated 
magnitude range for the Watts Bar site 
(between the New Madrid and the 
Charleston areas) . 

Figure 3.2.4d BEHCs which include only the 79 

contribution to the PGA hazard from 
earthquakes within the indicated 
magnitude range for the Catawba site 
(near the Charleston area). 

Figure 3.2.5 A plot of the relative location of all 80 

the sites in the study. The median 
PGA g-levels with a 10,000 year return 
period are also plotted. Rock sites 
are denoted by "R", deep soil sites by 
"D" and shallow soil sites by "S". 
The relative location of the New 
Madrid (NM) and Charleston (C) 
earthquakes are also shown. The 
circles indicate the highest hazard 
within an approximate regional area. 

Figure 3.2.6 Plot of the ratio of the probability of 81 

exceeding 0.3g PGA for the median (line), 
85th percentile (plot symbol, "0") and the 
arithmetic mean (plot symbol, "X") for the 
(shallow soil case)/(rock case). Site ID 
number is the same as the section number 
listed in Table 1.1. 

Figure 3.3.1 Comparison of the 10,000 year return 84 

period CPUHS between the Arkansas and 
Limerick sites (both rock sites). 



-xn- 



PAGE 



Figure 3.3.2 Comparison of the median 10,000 year 85 

return period CPUHS between the 
Seabrook, Catawba, Arkansas and Watts 
Bar sites (all rock sites). 

Figure 3.3.3 Comparison of the median 10,000 year 86 

return period CPUHS between two 
shallow soil sites (Till-like 2). The 
Clinton site (near New Madrid) to the 
Yankee Rowe site (New England). 

Figure 3.3.4 Comparison between the median 10,000 87 

year return period CPUHS for the sites 
with the lowest and highest median 
10,000 year CPUHS. 



-xm- 



Foreword 

The impetus for this study came from two unrelated needs of the Nuclear 
Regulatory Commission (NRC). One stimulus arose from the NRC funded "Seismic 
Safety Margins Research Programs" (SSMRP). The SSMRP's task of simplified 
methods needed to have available data and analysis software necessary to 
compute the seismic hazard at any site located east of the Rocky Mountains 
which we refer to as the Eastern United States (EUS) in a form suitable for 
use in probabilistic risk assessment (PRA). The second stimulus was the 
result of the NRC's discussions with the U.S. Geological Survey (USGS) 
regarding the USGS's proposed clarification of their past position with 
respect to the 1886 Charleston earthquake. The USGS clarification was finally 
issued on November 18, 1982, in a letter to the NRC, which states that: 

"Because the geologic and tectonic features of the Charleston region are 
similar to those in other regions of the eastern seaboard, we conclude 
that although there is no recent or historical evidence that other regions 
have experienced strong earthquakes, the historical record is not, of 
itself, sufficient ground for ruling out the occurrence in these other 
regions of strong seismic ground motions similar to those experienced near 
Charleston in 1886. Although the probability of strong ground motion due 
to an earthquake in any given year at a particular location in the eastern 
seaboard may be very low, deterministic and probabilistic evaluations of 
the seismic hazard should be made for individual sites in the eastern 
seaboard to establish the seismic engineering parameters for critical 
facilities." 

Anticipation of this letter led the Office of Nuclear Reactor Regulation to 
jointly fund a project with the Office of Nuclear Regulatory Research. The 
results were presented in Bernreuter et. al., (1985), and the objectives were: 

1. to develop a seismic hazard characterization methodology for the 
entire region of the United States east of the Rocky Mountains. 

2. to apply the methodology to selected sites to assist the NRC staff in 
their assessment of the implications in the clarification of the USGS 
position on the Charleston earthquake, and the implications of the 
occurrence of the recent earthquakes such as that which occurred in 
New Brunswick, Canada, in 1982. 

The methodology used in that 1985 study evolved from two earlier studies that 
the Lawrence Livermore National Laboratory (LLNL) performed for the NRC. One 
study, Bernreuter and Minichino (1983), was part of the NRC's Systematic 
Evaluation Program (SEP) and is simply referred hereafter to as the SEP 
study. The other study was part of the SSMRP. 

At the time (1980-1985), an improved hazard analysis methodology and EUS 
seismicity and ground motion data set were required for several reasons: 

Although the entire EUS was considered at the time of the SEP study, 
attention was focused on the areas around the SEP sites — mainly in 



-XIV- 



the Central United States (CUS) and New England. The zonation of 
other areas was not performed with the same level of detail. 

The peer review process, both by our Peer Review Panel and other 

reviewers, identified some areas of possible improvements in the SEP 
methodology. 

Since the SEP zonations were provided by our EUS Seismicity Panel in 
early 1979, a number of important studies had been completed and 
several significant EUS earthquakes had occurred which could impact 
the Panel members' understanding of the seismotectonics of the EUS. 

Our understanding of the EUS ground motion had improved since the 
time the SEP study was performed. 

By the time our methodology was firmed up, the expert opinions collected and 
the calculations performed (i.e. by 1985), the Electric Power Research 
Institute (EPRI) had embarked on a parallel study. 

We performed a comparative study, Bernreuter et. al., (1987), to help in 
understanding the reasons for differences in results between the LLNL and the 
EPRI studies. The three main differences were found to be: (1) the minimum 
magnitude value of the earthquakes contributing to the hazard in the EUS, (2) 
the ground motion attenuation models, and (3) the fact that LLNL accounted for 
local site characteristics and EPRI did not. Several years passed between the 
1985 study and the application of the methodology to all the sites in the 
EUS. In recognition of the fact that during that time a considerable amount 
of research in seismotectonics and in the field of strong ground motion 
prediction, in particular with the development of the so called random 
vibration or stochastic approach, NRC decided to follow our recommendations 
and have a final round of feedback with all our experts prior to finalizing 
the input to the analysis. 

In addition, we critically reviewed our methodology which lead to minor 
improvements and we also provided an extensive account of documentation on the 
ways the experts interpreted our questionnaires and how they developed their 
answers. Some of the improvements were necessitated by the recognition of the 
fact that the results of our study will be used, together with results from 
other studies such as the EPRI study or the USGS study, to evaluate the 
relative hazard between the different plant sites in the EUS. 

This report includes eight volumes: 

Volume I provides an overview of the methodology we developed for this 
project. It also documents the final makeup of both our Seismicity and 
Ground Motion Panels, and documents the final input from the members of 
both panels used in the analysis. Comparisons are made between the new 
results and previous results. 

Volumes II to V provide the results for all the active nuclear power plant 
sites of the EUS divided into four batches of approximately equal size and 
of sites roughly located in the four main geographical regions of the EUS 



-XV- 



{NE,1 SE, NC and SC) . A regional discussion is given in each of Vols II 
to V. 

Volume VI emphasizes important sensitivity studies, in particular the 
sensitivity of the results to correction for local site conditions and 
G-Expert 5's ground motion model. It also contains a summary of the 
results and provides comparisons between the sites within a common region 
and for sites between regions. 

Volume VII contains unaltered copies of the ten questionnaires used from 
the beginning of the 1985 study to develop the complete input for this 
analysis. 

After the bulk of the work was completed and draft reports for Vols. I-VII 
were written, additional funding became available. 

Volume VIII contains the hazard result for the 12 sites which were 
primarily rock sites but which also had some structures founded on shallow 
soil. These results supplement the results given in Vols. II to V where 
only the primary soil condition at the site was used. 



-XVI- 



List of Abbreviations and Symbols 

A Symbol for Seismicity Expert 10 in the figures displaying the results 
for the S-Experts 

ALEAS Computer code to compute the BE Hazard and the CP Hazard for each 
seismicity expert 

AM Arithmetic mean 

AMHC Arithmetic mean hazard curve 

B Symbol for Seismicity Expert 11 in the figures displaying the results 
for the S-Experts 

BE Best estimate 

BEHC Best estimate hazard curve 

BEUHS Best estimate uniform hazard spectrum 

BEM Best estimate map 

C Symbol for Seismicity Expert 12 in the figures displaying the results 
for the S-Experts 

COMAP Computer code to generate the set of all alternative maps and the 
discrete probability density of maps 

COMB Computer code to combine BE hazard and CP hazard over all seismicity 
experts 

CP Constant percentile 

CPHC Constant percentile hazard curve 

CPUHS Constant percentile uniform hazard spectrum 

CUS Central United States, roughly the area bounded in the west by the 
Rocky Mountains and on the east by the Appalachian Mountains, 
excluding both mountain systems themselves 

CZ Complementary zone 

D Symbol for Seismicity Expert 13 in the figures displaying the results 
for the S-Experts 

EPRI Electric Power Research Institute 



-XV n- 



EUS Used to denote the general geographical region east of the Rocky 

Mountains, including the specific region of the Central United States 
(CUS) 

g Measure of acceleration: Ig = 9.81m/s/s = acceleration of gravity 

G-Expert One of the five experts elicited to select the ground motion models 
used in the analysis 

GM Ground motion 

HC Hazard curve 

Iq Epicentral intensity of an earthquake relative to the MMI scale 

I5 Site intensity of an earthquake relative to the MMI scale 

LB Lower bound 

LLNL Lawrence Livermore National Laboratory 

M Used generically for any of the many magnitude scales but generally 
%, mfj{Lg), or M|_. 

Ml Local magnitude (Richter magnitude scale) 

Mjj True body wave magnitude scale, assumed to be equivalent to mj^CLg) 
(see Chung and Bernreuter, 1981) 

mfj(Lg) Nuttli's magnitude scale for the Central United States based on the 
Lg surface waves 

M5 Surface wave magnitude 

MMI Modified Mercalli Intensity 

Mq Lower magnitude of integration. Earthquakes with magnitude lower 
than Mq are not considered to be contributing to the seismic hazard 

NC North Central; Region 3 

NE North East; Region 1 

NRC Nuclear Regulatory Commission 

PGA Peak ground acceleration 

PGV Peak ground velocity 

PRO Computer code to compute the probability distribution of epicentral 
distances to the site 



-XVI 11- 



PSRV Pseudo relative velocity spectrum. Also see definition of spectra 
below 

Q Seismic quality factor, which is inversely proportional to the 
inelastic damping factor. 

Ql Questionnaire 1 - Zonation (I) 

Q2 Questionnaire 2 - Seismicity (I) 

Q3 Questionnaire 3 - Regional Self Weights (I) 

Q4 Questionnaire 4 - Ground Motion Models (I) 

Q5 Questionnaire 5 - Feedback on seismicity and zonation (11) 

Q6 Questionnaire 6 - Feedback on ground motion models (II) 

Q7 Questionnaire 7 - Feedback on zonation (III) 

Q8 Questionnaire 8 - Seismicity input documentation 

Q9 Questionnaire 9 - Feedback on seismicity (III) 

QIO Questionnaire 10 - Feedback on ground motion models (III) 

R Distance metric, generally either the epicentral distance from a 

recording site to the earthquake or the closest distance between the 
recording site and the ruptured fault for a particular earthquake. 

Region 1 (NE): North East of the United States, includes New England and 
Eastern Canada 

Region 2 (SE): South East United States 

Region 3 (NC): North Central United States, includes the Northern Central 
portions of the United States and Central Canada 

Region 4 (SC): Central United States, the Southern Central portions of the 
United States including Texas and Louisiana 

RP Return period in years. 

RY Random vibration. Abbreviation used for a class of ground motion 
models also called stochastic models. 

S Site factor used in the regression analysis for G-Expert 5's GM 
model: S = for deep soil, S = 1 for rock sites 

SC South Central; Region 4 

SE South East; Region 2 



-XIX- 



S-Expert One of the eleven experts who provide the zonations and seismicity 
models used in the analysis 

SEP Systematic Evaluation Program 

SHC Seismic Hazard Characterization 

SHCUS Seismic Hazard Characterization of the United States 

SN Site Number 

Spectra Specifically in this report: attenuation models for spectral 
ordinates were for 5% damping for the pseudo-relative velocity 
spectra in PSRV at five frequencies {25, 10, 5, 2.5, 1 Hz). 

SSE Safe Shutdown Earthquake 

SSI Soil-structure-interaction 

SSMRP Seismic Safety Margins Research Program 

UB Upper bound 

UHS Uniform hazard spectrum (or spectra) 

USGS United States Geological Survey 

WUS The regions in the Western United States where we have strong ground 
motion data recorded and analyzed 



-XX- 



Executive Summary: Volume VI 

Volume VI is one of eight volumes comprised in the reporting of this study. 
In this volume we discuss the important sensitivities found in the course of 
our analysis, we make regional comparisons between sites and summarize the 
conclusions we have reached. 

In Section 2 we give a discussion of the sensitivity of the computed seismic 
hazard to several important aspects of the methodology used to estimate the 
ground motion. First, in Section 2.2 we present the results of a sensitivity 
study to show the effect that site correction has on the hazard. We selected 
the Limerick site and performed the analysis assuming that the site fell into 
each of our eight soil categories. We then compared the results from these 
eight separate analyses. This comparison gives the effect of site type on the 
hazard at the Limerick site. In an effort to generalize from these results, 
we found three pairs of sites that are distant from the Limerick site but each 
pair are relatively close together and have different site categories. By 
comparing the effect of site type on the hazard observed at these sites to the 
effect of site type observed as part of our sensitivity study using the 
Limerick site, we reached the conclusion that there did not appear to be a 
significant variation in the effect of site type introduced by the region the 
site is located in. 

This point is re-examined in Vol. VIII and summarized in Section 2 of Vol. VI 
where it is confirmed that the amount of variation from the expected value 
varies with the regional location of the site. However, this amount of 
variation was not found to be greater than 10 percent at the 12 sites 
analyzed. We also emphasize the dangers of reasoning in terms of 
probabilities of exceedance, since comparable variations in PGA between two 
sites could translate into drastically different amounts of variation in the 
probabilities of exceedance, due to the different slopes of the hazard curves. 

In Section 2.3 we revisited the sensitivity of the results to the inclusion or 
non-inclusion of G-Expert 5's ground motion model. We identified four 
categories of sites: (1) rock with the hazard from distant zones with large 
earthquakes, (2) rock with the hazard primarily from local zones, (3) same as 
(1) except a soil site, and (4) same as (2) except a soil site. We found that 
the results were most sensitive to the inclusion or non-inclusion of G- 
Expert's 5 ground motion model for category (1), then followed by decreasing 
sensitivity for category (2), then category (3) and least sensitivity for 
category (4). Interestingly, we found the sensitivity of the median to the 
inclusion/non-inclusion of G-Expert 5's model was about the same for all four 
categories of sites. 

In Section 2.4 we examined the reasons why our constant percentile uniform 
hazard spectra seemed to be high relative to the hazard curve for PGA. We 
concluded that the apparent disconnect between the PGA hazard and the spectral 
hazard was due to the correction for EDS conditions introduced into some of 
the ground motion models. These corrections suggest that typical EUS 
earthquakes have significantly more high frequency motion than assumed in 
either the R.G. 1.60 spectrum or the NUREG-0098 spectrum. 



-xxi- 



In Section 3 we compare the results between all sites. At the 0.2g level we 
found that typically at any site there is over two orders of magnitude 
uncertainty in the estimate of hazard (as measured by the difference between 
the 15th and 85th percentiles CPHCs) . We also found that the spread of the 
median probability of exceeding 0.2g PGA between the site with the lowest 
hazard and the site with the highest hazard is about 1.4 orders of magnitude. 

We did not find large differences in the hazard between sites located at 
approximately the same distance (approximately 200 km) from the New Madrid 
seismic zone, as with sites located approximately the same distance from the 
Charleston seismic zone. We did find that the makeup of the hazard was 
different with nearby zones being more important for sites near the Charleston 
seismic zone than for sites near the New Madrid zones. Conversely, large 
distant earthquakes were more important for the New Madrid site than for the 
Charleston site. 

We found that, of the sites analyzed , some sites in New England had the 
highest hazard. But it must be noted that the sites affected by the New 
Madrid and Charleston earthquake were at some distance from the source 
zones. Thus, if a site were to be located near these source zones, the hazard 
would be greater than found for the New England sites. 

Some regional influence could be seen in the spectral shape, particularly at 
the longer periods. The spectral shapes for the sites near the New Madrid 
region had more long period energy than for sites located near Charleston or 
in New England. There were some differences at the short period end of the 
spectrum, but it was relatively small. 

In Section 4 we present a number of conclusions reached during the course of 
this study, the most important of which are: 

Our estimates of the seismic hazard for any site in the EUS have 
large uncertainties at some sites. Most individual experts have 
expressed significant uncertainties about their input. There is also 
a wide diversity in the opinion among experts. 

The median estimate of the seismic hazard appears to be relatively 
stable, both in time and between studies performed without systematic 
differences. 

Correction for local soil conditions is important and has a 
significant impact on the results. 

The results, particularly the arithmetic mean and best estimate 
estimators, are very sensitive to ground motion models with low 
attenuation, e.g., such as the model selected by G-Expert 5. 

There is a significant variation in the hazard across the EUS; e.g., 
the median estimate for the 10,000 year return period for the PGA 
varies from 0.08g to 0.33g. 



-xxn- 



1. INTRODUCTION 

In Vol. I of this report, we provide a discussion of our methodology and the 
input provided by both our S and G-Experts. In Vols. II-V, we provided the 
results of our analysis of the seismic hazard for all of the active Eastern 
United States (EUS) nuclear power plant sites using our methodology and the 
input from our experts. 

In writing this volume (Vol. VI) we have assumed that the reader had read 
Vol. I and any one of Vols. II-V. In Section 2 of this volume we give a 
discussion of the sensitivity of the results presented in Vols. II-V to 
several important aspects of the methodology we used to estimate the ground 
motion. 

In each of the Vols. II-V we included a section where comparisons were made 
between the sites included in the volume. In Section 3 of this volume we make 
comparisons between all sites and regions. In Section 4 we summarize the 
conclusions we have reached at the end of this study. 

Volume VII gives all of the questionnaires answered by our experts. The 
experts responses to these questionnaires form the basis for input data needed 
by our computer programs to compute the seismic hazard at each site. 

Volume VIII provides additional analysis to account for the hazard at sites 
where critical structures are founded on several soil categories. 

As discussed in Vol. I, the 69 active nuclear power plant sites were divided 
into four batches - roughly along regional lines. However, in order to have 
approximately four equal sets, the sites in batch 4 did not correspond to a 
single region. In Tables 1.1a to l.ld we list all of the sites included in 
each of the Volumes of this report and in Figs. 1.1a, b, c, d we provide maps 
giving the location of each of the sites. In Fig. 1.2 we provide a figure 
with all of the sites (unlabeled) located on it to establish the relative 
location between the sites and other points of interest. 

In Section 2 of this Volume we examine the sensitivity of the computed seismic 
hazard to several important aspects of the methodology used to estimate the 
ground motion. In previous Volumes we have indicated the significance of 
correction introduced to model the local soil conditions at various sites. In 
Section 2.2 we give the results of a sensitivity study of the effect of our 
methodology for site correction on the hazard estimates. In addition, we 
provide a discussion on the amount of ground motion amplification in the PGA 
values between shallow and rock site conditions at the 12 sites which include 
both rock and shallow soil conditions (see Vol. VIII). 

In previous Volumes we have noted the sensitivity of the hazard at some rock 
sites to the ground motion model selected by G-Expert 5. In Section 2.3 we 
explore this issue in detail. 

In Section 2.4 we discuss the implication on the hazard of the correction that 
has been introduced into some ground motion models heavily weighted by G- 
Experts to "correct" them for use in the EUS. 



-1- 



In Section 3 we compare the estimated hazard between sites and discuss 
regional variations in the estimated hazard. 

In Section 4 we summarize the main results and conclusions reached in this 
study. 



-2- 



TABLE 1.1a 

SITES AND SOIL CATEGORY USED FOR EACH SITE 
IN BATCH 1 







MAP (1) 






SITE NAME 


KEY 


SOIL CATEGORY (2) 


1. 


Fitzpatrick 


1 


Rock 


2. 


Ginna 


2 


Rock 


3. 


Haddam Neck 


3 


Rock 


4. 


Hope Creek 


4 


Deep Soil 


5. 


Indian Point 


5 


Rock 


6. 


Limerick 


6 


Rock 


7. 


Maine Yankee 


7 


Rock 


8. 


Millstone 


8 


Rock 


9. 


Nine Mile Pt. 


9 


Rock ** 


10. 


Oyster Creek 


A 


Deep Soil 


11. 


Peach Bottom 


B 


Rock 


12. 


Pilgrim 


C 


Sand-Like 2 


13. 


Salem 


D 


Deep Soil 


14. 


Seabrook 


E 


Rock 


15. 


Shoreham 


F 


Deep Soil 


16. 


Susquehanna 


G 


Rock** 


17. 


Three Mile Island 


H 


Rock** 


18. 


Vermont Yankee 


I 


Rock 


19. 


Yankee at Rowe 


J 


Till- Like 2 



(1) Key used on Fig. 1.1a 

(2) Site categories are given in Table 3.9 of Vol. I and repeated in 
Table 1.2. 

(**) Have some structures founded on shallow soil. 



-3- 



TABLE 1.1b 

SITES AND SOIL CATEGORY USED FOR EACH SITE 
IN BATCH 2 





SITE NAME 


1. 


Bellefonte 


2. 


Browns Ferry 


3. 


Brunswick 


4. 


Calvert Cliffs 


5. 


Catawba 


6. 

7. 


Farley 
Hatch 


8. 


McGuire 


9. 


North Anna 


10. 


Oconee 


11. 


Robinson 


12. 
13. 


Sequoyah 
Shearon Harris 


14. 


Sumner 


15. 
16. 
17. 


Surry 
Vogtle 
Watts Bar 



Map (1) 
KEY 

1 
2 
3 
4 
5 
6 
7 
8 
9 
A 
B 
C 
D 
E 
F 
G 
H 



SOIL 


CATEGORY 


Rock 




Rock 


** 


Till- 


-like 2 


Deep 


soil 


Rock 


** 


Rock 


** 


Deep 


soil 


Rock 




Rock 


** 


Rock 


** 


Deep 


soil 


Rock 




Rock 




Rock 


** 


Deep 


soil 


Deep 


soil 


Rock 





(1) Key used on Fig. 1.1b. 

(**) Have some structures founded in shallow soil 



-4- 



TABLE 1.1c 

SITES AND SOIL CATEGORY USED FOR EACH SITE 
IN BATCH 3 





SITE NAME 


1. 


Beaver Valley 


2. 


Big Rock Point 


3. 


Braidwood 


4. 


Byron 


5. 


Clinton 


6. 


Cook 


7. 


Davis Besse 


8. 


Dresden 


9. 


Fermi 


10. 


Kewaunee 


11. 


LaSalle 


12. 


Palisades 


13. 


Perry 


14. 


Point Beach 


15. 


Quad Cities 


16. 


Zion 



Map (1) 
KEY 

1 
2 
3 
4 
5 
6 
7 
8 
9 
A 
B 
C 
D 
E 
F 
G 



SOIL CATEGORY 



Sand-1 ike 

Till-like 

Rock 

Rock 

Till-like 

Sand-like 

Rock 

Rock 

Rock 

Till-like 

Till-like 

Sand-like 

Rock 

Till-like 

Rock 

Sand-like 



(1) Key used on Fig. 1.1c. 

(**) Have some structures founded in shallow soil 



-5- 



TABLE l.ld 

SITES AND SOIL CATEGORY USED FOR EACH SITE 
IN BATCH 4 

MAP (1) 





SITE NAME 


1. 


Arkansas 


2. 


Callaway 


3. 


Comanche Peak 


4. 


Cooper 


5. 


Crystal River 


6. 


Duane Arnold 


7. 


Fort Calhoun 


8. 


Grand Gulf 


9. 


LaCrosse 


10. 


Monticello 


11. 


Prairie Island 


12. 


River Bend 


13. 


South Texas 


14. 


St. Lucie 


15. 


Turkey Point 


16. 


Waterford 


17. 


Wolf Creek 



KEY 



1 
2 
3 
4 
5 
6 
7 
8 
9 
A 
B 
C 
D 
E 
F 
G 
H 



SOIL CATEGORY 



*• 



** 



Rock 
Rock 
Rock 

Sand-like 
Rock 
Rock ** 
Sand-like 
Deep soil 
Sand-like 
Sand- like 
Sand-like 
Deep soil 
Deep soil 
Deep soil 
Rock 

Deep soil 
Rock 



(1) Key used on Fig. l.ld. 

(**) Have some structures founded in shallow soil 



-6- 




Figure 1.1a Map showing the location of the Batch 1 sites contained in 

Vol. II of this report. Map symbols are given in Table 1.1a, 



-7- 




Figure 1.1b Map showing the location of the Batch 2 sites contained in 

Vol. Ill of this report. Map symbols are given in Table 1.1b. 



-8- 




Figure 1.1c Map showing the location of the Batch 3 sites contained in 
Vol. IV of this report. Map symbols are given in Table 1.1c 



-9- 




Figure l.ld Map showing the location of the Batch 4 sites contained in 
Vol. V of this report. Map symbols are given in Table l.ld 



-10- 




Figure 1.2 Map giving the relative location of all the sites included in 
this study. 



-11- 



2. DISCUSSION OF THE SENSITIVITY OF THE COMPUTED 

SEISMIC HAZARD TO SEVERAL IMPORTANT ASPECTS OF 
THE METHODOLOGY USED TO ESTIMATE THE GROUND MOTION 

2.1 Background 

In Vols. II-V we have identified several important elements of the methodology 
used to estimate the ground motion that have a significant impact on the 
results. One of the most significant elements is the method we used to 
correct the estimated ground motion to account for the site's soil category. 
A number of examples of the significance of this correction have been pointed 
out in Vols. II-V. In Section 2.2 we provide a detailed examination of the 
impact of this element of our methodology on the estimated hazard. 

In Vols. II-V we found that differences in the rate at which the ground motion 
attenuated between the various ground motion models selected by our G-Experts 
led, in some cases, to relatively large uncertainties in the estimated 
hazard. In particular the ground motion model selected by 6-Expert 5 
sometimes lead to hazard estimates significantly higher than from the other 
ground motion models for a given S-Expert's input. This element is further 
discussed in Section 2.3. 

In Section 2.4 we discuss an element not previously mentioned. Based on a 
number of studies, e.g., Newmark and Hall (1978), of the relation between PGA 
and the spectral amplification factors, it is generally assumed that at 
approximately 33hz the spectral amplification of the PGA is unity. However, 
if the estimated peak ground acceleration (PGA) hazard given in Vols. II-V is 
converted to a high frequency spectral value and compared to appropriate 
spectrum, our PGA values appear to be anomalously low, or conversely our 
spectra would suggest that at high frequency (25hz and above) we have higher 
amplification factors than generally assumed based on spectra from WUS 
earthquakes or design spectra such as NUREG-0098. 

2.2 Correction for the Site's Soil Category 

In Vol. I Section 3.7 we provide a discussion of the overall approach we used 
to account for the site's soil category. In Bernreuter et al. (1987) we 
performed a sensitivity study on site type. However, that study was based on 
the early input provided by our G-Experts and documented in Bernreuter et al . 
(1985). The results in this section are significantly different from our 
previous results primarily because in the previous results a site correction 
was only applied about half the time. In the updated input four out of the 
five G-Expert selected our category approach with a weight of 1.0. The other 
G-Experts selected the simple correction approach with three soil classes: 
rock, shallow soil and deep soil. In Bernreuter et al. (1985 & 1987) we only 
allowed for rock and deep soil classes. The net result of the changes in 
input provided by our G-Experts and our expanded capabilities of allowing for 
more sub-divisions in the simple correction approach is that the site's soil 
category is now much more significant than previously found in Bernreuter et 
al. (1985 & 1987). 



-12- 



As explained in Vol. I Section 3.7 we put all sites in one of the eight soil 
categories listed in Table 2.2.1. For the four 6-Experts who selected our 
categorized approach explained in Vol. I for each trial of the Monte Carlo 
simulation a site correction is simulated assuming that the distribution of 
the correction factor is lognormal with a median plotted in either Fig. 2.2.1 
or Fig. 2.2.2. These figures are repeated from Vol. 1 for ease of 
reference. For 6-Expert 5 a constant correction value was always used based 
on the site's soil category. See Fig. 2.2.3. 

To see the significance of the soil category we randomly selected a site and 
re-ran the analysis eight times with the site's soil category changed each 
time. In Fig. 2.2.4 we compare the CPHCs between the case when the site's 
soil category is assumed to be rock and the case when it is assumed to be deep 
soil. The differences between the median CPHCs for the rock and deep soil 
categories is relatively small. However, as shown in Fig. 2.2.5 there are 
much larger differences between the two cases for other estimators. In 
particular, as can be seen from Fig. 2.2.5, the largest difference is between 
BEHCs for the two cases. It should be noted that there is a significant 
variation in the AMHC and BEHC from region to region. Thus the differences 
between BEHCs for the rock and soil cases and the differences between AMHCs 
observed cannot be generalized and assumed to occur at other sites. For this 
reason we primarily focus on the CPHCs and CPUHS. 

In Fig. 2.2.6 we compare the CPUHS with a 10,000 year return period for the 
case when the site's soil category is deep soil to the case when it is rock. 
We see from Fig. 2.2.6 that the spectral shapes are significantly different. 
Note, that the difference would appear to be much larger if the spectral 
velocity scale was linear (as the PGA scale in the preceding figures) rather 
than logarithmic. 

In Fig. 2.2.7 we compare the CPHCs between the cases when the site's soil 
category is considered as: (1) Till-1, (2) Till-2 and (3) Till-3. Also, 
shown for reference is the case when the site's soil category is considered to 
be rock. We see from Fig. 2.2.7 that, as might be expected from examination 
of the median correction factors in Fig. 2.2.1, there is a significant 
difference between the hazard curves depending upon the site's soil 
category. In Fig. 2.2.8 we compare the median 10,000 yr return period CPUHS 
for the cases when the site's soil category is: (1) Till-l, (2) Till-2, (3) 
Till-3 and, for reference, (4) rock. Note the change in spectral shape with 
site category. Once again keep in mind that a logarithmic scale is used. In 
Fig. 2.2.9 we include both the 15th and 85th percentile 10,000 yr return 
period CPUHS as well. 

In Fig. 2.2.10 we compare the CPHCs for the cases when the site's soil 
category is considered to be: (1) Sand-1, (2) Sand-2, (3) Sand-3 and, for 
reference, (4) rock. In Fig. 2.2.11 we compare the median 10,000 yr return 
period CPUHS for the cases when the site's soil category is considered to be: 
(1) Sand-1, (2) Sand-2, (3) Sand-3 and, for reference, (4) rock. In 
Fig. 2.2.12 we include both the 15th and 85th percentile 10,000 yr return 
period CPUHS for the cases used in Fig. 2.2.11. 

-13- 



We see by comparing Figs. 2.2.7 to 2.2.10 for PGA and Figs. 2.2.8 to 2.2.11 
for spectra that at the short period end of the spectrum there is little 
difference in the hazard due to the site's soil category between the Till-1 or 
Sand-1 categories and a somewhat larger difference between the Til 1-2 and 
Sand-2, and between the Til 1-3 and Sand-3. In all cases the Till-like hazard 
is lower. We also see from comparing Fig. 2.2.8 to Fig. 2.2.11 there are some 
differences in spectral shape between the Till-like categories and the Sand- 
like categories. 

The above comparisons {Figs. 2.2.4 to 12) are for a fixed site. It is natural 
to ask if the differences are independent of regional seismicity. This can be 
in part addressed using comparisons made in Vols. II-V. First, it should be 
noted that Figs. 2.2.4 to 12 were based on the zonation and seismicity for the 
Limerick site. Thus, to see if the differences between hazard curves due to 
site's category is regionally independent we need adjacent (or almost 
adjacent) sites with different site categories located relatively far (at 
least 200 km) from the Limerick site (see Fig. 1.1a for the location of the 
Limerick site). The difficult issues in selecting pairs of sites with 
different soil categories that we can use to address the potential regional 
variation of the relative change in the hazard at sites due to the site's soil 
category is to have a criteria for how "close" the sites must be together. We 
have seen in Vols. II-V that there can be some significant variations between 
the hazard for relatively nearby sites due to zonation differences. In 
Table 2.2.2 we list three pairs of sites which, in our opinion, are 
sufficiently close together and have different soil categories. 

In Fig. 2.2.13a we compare the median CPHCs between the Vermont Yankee site 
(rock) and the Yankee Rowe site (Til 1-2). In Fig. 2.2.13b we compare the 
median CPHCs between the Braidwood (rock) and the LaSalle (Til 1-2) sites and 
in Fig. 2.2.13c we compare the median CPHCs between the Kewaunee (Til 1-2) and 
Point Beach (Till-1) sites. We see by comparing the relative difference 
between the Till-1 median hazard curves between the rock and Til 1-2 sites in 
Figs. 2.2.7, 2.2.13a and 2.2.13b that there is some regional variation. We 
see that the relative differences between the rock and Till-2 curves is 
approximately the same between Figs. 2.2.7 and 2.2.13a. However, there is a 
slightly larger spread between the median rock and Till-2 hazard curves in 
Fig. 2.2.13b than in Figs. 2.2.7 and 2.2.13a at the high g value end. The 
spread between the curves is approximately 20% larger, in terms of annual 
probability of exceedance at PGA levels greater than Ig in Fig. 2.2.13b than 
in Figs. 2.2.7 and 2.2.13a. At the low g value end there is very little 
difference in the relative spread between the curves. One reason why the 
relative effect of site category varies from site to site is because, as 
discussed in Vols. II-V, the uncertainty is larger at some sites than at 
others; hence there is a larger variability between the estimated median 
between two different Monte Carlo simulations. In particular, this 
variability is relatively large at the Braidwood site. Reference should be 
made to Fig. 3.1.1 in Vol. IV. The regional variation in the effect of site 
correction on the hazard is examined in some detail in Vol. VIII. It shows 
that there can be a large variation between the correction for site type and 
in particular either the AMHC or the 85th percentile CPHC. This is 

-14- 



illustrated in Fig. 2.2.14 where we compare the CPHC for the Browns Ferry site 
for the case when it is ran as a deep soil site to the case when it is ran as 
a rock site. We see that there is about the same variation between the median 
CPHCs for the two cases as shown in Fig. 2.2.4, however, there is a much 
larger difference between the 85th percentile CPHCs for the two cases in Fig. 
2.2.14 than in Fig. 2.2.4. The same difference is observed between the AMHCs 
for the two cases for the two sites as can be seen by comparing Figs. 2.2.5 to 
2.2.15. 

It should be noted based on the preceding comparisons that although there 
appears to be little regional variation in the correction for site type for 
the median hazard curve there is, as discussed in Section 3, considerable 
regional variation in the resultant spectral shape for a given soil category. 

If the median correction factors given in Figs. 2.2.1 and 2.2.2 are compared 
to the comparisons made between median CPHC for PGA and median CPUHS, one 
finds for the sites studied that the simulated median hazard curve or UHS for 
a given site category is within approximately 10-20 percent of the rock case 
multiplied (parallel to the PGA axis) by the appropriate median correction 
factor. See Section 3 for details. Generally the simulated curve is lower 
than the curve obtained by the simple correction procedure outlined above. 
For sites with large uncertainties there can be larger variation, however, 
this is due to the variation between successive Monte Carlo runs. At higher 
probability of exceedances the difference between simulated correction and a 
simple ratio correction is larger than at low probabilities of exceedance. It 
is, however, difficult to estimate the impact of soil variation on either the 
85th percentile CPHCs or AMHCs. 

In Volume VIII we provide calculations for the seismic hazard at the current 
nuclear power plant sites which have critical structures founded on several 
soil categories. In Volume VIII it is concluded that, in general, it is not 
correct to account for site effects simply by scaling the final hazard 
curve. There are cases where it would be correct, but at the present time 
there is no simple way of knowing which are these cases without performing a 
full analysis. 



-15- 



TABLE 2.2.1 
DEFINITION OF THE EIGHT SITE CATEGORIES 

CATEGORY DEPTH 

Generic Rock 

(1) Rock N/A 

Sand Like 



(2) Sand 1 

(3) Sand 2 

(4) Sand 3 

Till-Like 

(5) Till 1 

(6) Till 2 

(7) Till 3 



Deep Soil 

(8) Deep Soil N/A 



SI 


25 to 80 ft. 


S2 


80 to 180 ft. 


S3 


180 to 300 ft 



Tl 


25 to 80 ft. 


T2 


80 to 180 ft. 


T3 


180 to 300 ft 



•16- 



TABLE 2.2.2 
NEARBY SITES IN DIFFERENT SOIL CATEGORIES 



Site Pair 


Soil Category 


Location Map 
Fig. Number 


Vermont Yankee 
Yankee Rowe 
North East U.S. 


Rock 
Till-2 


1.1a 
1.1a 


Braidwood 

LaSalle 

North Central U.S. 


Rock 
Till-2 


1.1c 
1.1c 


Kewaunee 

Point Beach 

North Central U.S. 


Till-2 
Till-1 


1.1c 
1.1c 



-17- 



3.0j- 
2.8- 
2.8- 
2.* - 
2.2- 
2.0- 

o 

£ u - 

i 

I u- 

O 

« 12 - 

1.0 - - 
0.8 - - 
0.6- 
0.4- 
0.2- 



0.0 



SuViiL 


Category 


• 


T-1 


■ 


T-2 


♦ 


T-3 




0.001 



0.010 



0.100 
P«rlod, (tc 



1000 



10.000 



Figure 2.2.1 



smoothed median correction factors for the Till-like ^atagories 
listed in Table 2.2.1 relative to rock. The PGA correction 
factors are plotted at 0.01s. 



3.0 



2.8- 

2.6- 

2.4 

i2-- 

2.0- 

S t8-- 
o 

* 1.6- - 

c 
o 

i 1.* 

o 
" 1.2 

to-- 

0.8- 
0.6- 
0.4- 
0.2- 



0.0 



_Curve__Categor^ 
• Deep Soil 



S-1 
S-2 
S-3 




0.001 



0.100 
Ptrlod, t«e 



tooo 



10.000 



Figure 2.2.2 



Smoothed median correction factors for the Sand-like categories 
listed in Table 2.2.1 relative to rock. The PGA correction 
factors are plotted at 0.01s. 



-18- 



1000 
3.0- 



100 



Frequency, hentz 
10 



0.10 



2.5- 



u 
o 

t- 



2.0 



.« 1.5 + 



c 
o 



a. 

E 



-I c 1 — I — I I I I 



-I— 1 1 1 — I I I I I 



-I 1 I I i — I I I I 



I I I I I — I I I 



Tnifunac-Anderson 
Acceleration 
Factors. 1977 



1.0-- ♦Hock 



0.5-- 



0.0 



I Interndlats 



IDaep Soil 



_L. 



0.001 



0.010 



^ 



Trifunac-Lee. 1985 

^•Oaep Soli 

-Ulntenndlate 

-♦flock 



,- • ^ 



_L 



_L 



0.100 
Period, sec 



1.000 



10.000 



Figure 2.2.3 Simple correction factors selected by G-Expert 5. 



-19- 



E.U.S SEISMIC HAZARD CHARACTERIZATION 
LOWER MAGNITUDE OF INTEGRATION IS 5.0 

PERCENTILES = 15., 50. AND 85. 



< 

UJ 

>- 
q: 

UJ 



^ 



-1 
10 



-2 
10 



-3 
10 



HAZARD CURVES USING ALL EXPERTS 



ui -4 
o 10 



H -5 
-J 10 

CD 
< 
CD 
O 
CH 

a. 



-6 

10 



-7 

10 




ACCELERATION CM/SEC* '2 

LIMERICK 



Figure 2.2.4 



Comparison between the CPHCs for the case when the Limerick 
site's soil category is rock and the case when it is considered 
to be deep soil. 



-20- 



E.U.S SEISMIC HAZARD CHARACTERIZATION 
LOWER MAGNITUDE OF INTEGRATION IS 5.0 



-1 
10 



-7 
10 



HAZARD CURVES USING ALL EXPERTS 



I I I 



-BEST.A-ARITHMETIC 




•- <N CM 



ACCELERATION CM/SEC»»2 

LIMERICK 



Figure 2.2.5 



Comparison between the AMHCs and the BEHCs for the case when 
the Limerick site's soil category is rock and the case when it 
is deep soil. 



-21- 



E.U.S SEISMIC HAZARD CHARACTERIZATION 
LOWER MAGNITUDE OF INTEGRATION IS 5.0 

10000. -YEAR RETURN PERIOD CONSTANT PERCENTILE SPECTRA FOR 

PERCENTILES = 15.. 50. AND 85. 



o 

10 



o 



u 
o 




to •«* 1/^ lOI^^OOtD 



I o 



PERIOD (SEC) 2 

LIMERICK 



Figure 2.2.6 



Comparison between the CPUHS with a 10,000 year return period 
for the case when the Limerick site's soil category is rock and 
the case when it is deep soil. 



-22- 



E.U.S SEISMIC HAZARD CHARACTERIZATION 
LOWER MAGNITUDE OF INTEGRATION IS 5.0 

PERCENTILES = 15., 50. AND 85. 



oe. 

< 



o 

X 

LJ 



O 

V 



-1 
10 



-2 
10 



HAZARD CURVES USING ALL EXPERTS 



UJ o 
Q. 10 



o 

z 

UJ 



-4 

10 



-I 10 



ffi 

< 
m 
o 
q: 

Q. 



-6 
10 



-7 
10 




ACCELERATION CM/SEC* '2 



LIMERICK 



Figure 2.2.7 



Comparison between the CPHCs for the case when the Limerick 
site's soil category is considered to be: Till-1 , Till-2 and 
Till-3. For comparison the rock case is also plotted. 



-23- 



E.U.S SEISMIC HAZARD CHARACTERIZATION 
LOWER MAGNITUDE OF INTEGRATION IS 5.0 



50-TH PERCENTILE SPECTRA FOR ALL RETURN PERIODS 



10 



2 
10 I- 



o 

UJ 



S ,o' 



o 
o 



10 



-1 
10 



CM 

I o 



I 1 1 1 
Rock 



- Till 1 



■^ Till 2 
-■ Till 3 




^o -^ intDt^^ooj) 



I o 



PERIOD (SEC)°o 
LIMERICK 






Figure 2.2.8 



Comparison between the median 10,000 year reutrn period CPUHS 
for the case when the Limericks site soil category is 
considered to be: Till-1 , Till-2 and Till-3. For comparison 
the rock case is also plotted. 



-24- 



E.U.S SEISMIC HAZARD CHARACTERIZATION 
LOWER MAGNITUDE OF INTEGRATION IS 5.0 

10000. -YEAR RETURN PERIOD CONSTANT PERCENTILE SPECTRA FOR 

PERCENTILES = 15. . 50. AND 85. 



10 



u 

UJ 

to 



o 



10 



10 



10 



-1 
10 



I o 




^■■ta^i^ita 



to •^ iDlOrvOQj) 



I o 



PERIOD (SEC) 2 

LIMERICK 



fO -^ iTHDf'JXXy' 



Figure 2.2.9 



Same as Fig. 2.2.8 except the 15th and 85th percentile CPUHS 
are also plotted. 



-25- 



E.U.S SEISMIC HAZARD CHARACTERIZATION 
LOWER MAGNITUDE OF INTEGRATION IS 5.0 

PERCENTILES = 15.. 50. AND 85. 



a: 

< 

>- 
on 

UI 
Q. 



< 

o 

UI 
UJ 

o 

X 

UI 



CD 
< 

m 
o 
a: 
a. 



-1 
10 



-2 
10 



-3 

10 



10 



10 



-6 

10 



-7 
10 



HAZARD CURVES USING ALL EXPERTS 




O 



K) •* in (o 

ACCELERATION CM/SEC* »2 

LIMERICK 



Figure 2.2.10 Comparison between the CPHCs for the cases when the Limerick 
site's soil category is considered to be: Sand-1 , Sand-2 and 
Sand-3. For comparison the rock case is also plotted. 



-26- 



E.U.S SEISMIC HAZARD CHARACTERIZATION 
LOWER MAGNITUDE OF INTEGRATION IS 5.0 



50-TH PERCENTILE SPECTRA FOR ALL RETURN PERIODS 



10 



o 
o 



10 



o 

Ul 

> 1 

^ 10 

>■ 



10 



-1 

10 



"'^T 



CM 
I O 



- Rock 

-• Sand 1 

^ Sand 2 

■m Sand 3 




To 



CM (O -<t irtlDhsOaj) 

PERIOD (SEC) 2 

LIMERICK 



K) •* in iDrvji3cr> 



Figure 2.2.11 Comparison between the median 10.000 year return period CPUHS 
for the cases when the Limerick site's soil category is 
considered to be: Sand-1 , Sand-2 and Sand-3. For comparison, 
the rock case is also plotted. 



-27- 



E.U.S SEISMIC HAZARD CHARACTERIZATION 
LOWER MAGNITUDE OF INTEGRATION IS 5.0 

10000. -YEAR RETURN PERIOD CONSTANT PERCENTILE SPECTRA FOR 

PERCENTILES = 15.. 50. AND 85. 



10 



o 

LJ 

o 

>- 



o 
o 



10 



10 



10 



-1 
10 



I O 



- Rock 

-• Sand 1 

■A Sand 2 

m Sand 3 




to -^ IT) JOr^OOT) 

To 



I I I I 



PERIOD (SEC) °o 

LIMERICK 






Figure 2.2.12 Same as Fig. 2.2.11 except the 15th and 85th percentile CPUHs 
are also plotted. 



-28- 



-1 

10 



-7 
10 



Vermont Yankee 
Yankee Rowe 




— <N CM 
O 

+ 



ACCELERATION CM/SEC* '2 



Figure 2.2.13a Comparison of the median CPHCs between the Vermont Yankee site 
(rock) and the nearby Yankee Rowe site (Till-2). 



-29- 



-1 

10 



,-^ 


-7 


< 


10 


Ui 




>- 




or 




LJ 




Q. 










-i 


UI 


10 


z 




< 




o 




LlI 




LJ 




O 




X 




UI 


-4 


u 


10 


O 




>- 




(- 




_J 




CO 

< 


-5 


ffi 


10 


O 




q: 




Q. 






-6 




10 



-7 
10 



B ra I dwood 
LaSalle 




o 

+ 

UJ 



ACCELERATION CM/SEC* '2 



Figure 2.2.13b Comparison of the Median CPHCs between the Braidwood site 
(rock) and the nearby LaSalle site (Till-2). 



-30- 



< 



a. 



UJ 

u 

z 

Ul 
Ul 

u 



CO 

< 

00 

o 

ce. 
a. 



-1 
10 



-2 
10 t 



-3 
10 



-4 
10 



-5 
10 



-6 

10 It 



-7 
10 



1 - 



z - 



Kewaunee 
Point Beach 




r- CN CM 

o 



ro -^ in lo r^ 00 
ACCELERATION CM/SEC* *2 



Figure 2.2.13c Comparison between the Median CPHCs for the Kewaunee site 
(Till-2) and the Point Beach site (Till-1). 



-31- 



-1 

10 



10 
< 

UI 

>- 

^ -."^ 

Q. 10 



u 

z 
< 

UI ^ 

o 10 

X 



o 



LOWER MAGNITUDE OF INTEGRATION IS 5.0 

PERCENTILES = 15., 50. AND 85. 



HAZARD CURVES USING ALL EXPERTS 



CD 
< 
CD 

o 

a: 
a. 



10 



-6 
10 



10 




CM fo ■* in 
ACCELERATION CM/SEC 



BROWNS FERRY 



Figure 2.2.1M Comparison between the CPHCs for case when the Browns Ferry 
site is rock and the case when it is treated as a deep soil 
site. 



-32- 



LOWER MAGNITUDE OF INTEGRATION IS 5.0 



-1 
10 



-7 
10 



HAZARD CURVES USING ALL EXPERTS 



B-BEST,A-ARITHMETIC 




X 



o 

+ 



■* in (o i~~ 

ACCELERATION CM/SEC*»2 

BROWNS FERRY 



Figure 2.2.15 Comparison between the AMHCs and BEHCs for the Browns Ferry 
site ran as a rock site and as a deep soil site. 



-33- 



2.3 Sensitivity to G-Expert 5's Model 

In Vols. II-V we have noted (e.g., Vol. II Sec. 2.1) that at rock sites 
G-Expert 5's BEHC per S-Expert is significantly higher than the other 
G-Experts' BEHCs per S-Expert. See Fig. 2.3.1a for a typical example. In 
some instances, particularly for the case when a site is several hundred 
kilometers of a zone such as the New Madrid seismic zone and located for a 
given S-Expert in a zone of very low seismic activity (i.e., where the 
dominant contribution comes from a distant zone more than 150 to 200 km away, 
with high magnitude cutoff) , there can be a much larger difference between G- 
Expert 5's BEHC per S-Expert and the other G-Experts' BEHCs for the same S- 
Expert as shown in Fig. 2.3.1b. 

Although we have discussed this issue at some length in each of Vols. II-V, it 
is sufficiently important to warrant additional discussion. As discussed in 
Vols. II-V the main reasons for the large spread between G-Expert 5's BEHC per 
S-Expert and the other G-Experts' BEHCS per the same S-Expert are: 

(1) As discussed in Vol. 1 Section 3.5, the site correction factor 
selected by G-Expert 5 at rock sites multiplies the PGA by 
approximately a factor of 2 relative to the other BE GM models for 
the same distance and magnitude. A factor of 2 in PGA results in 
approximately a factor of 8-10 (or even more at low PGA values) in 
probability of exceedanceT 

(2) It can be seen from Fig. 3.4 in Vol. 1 that G-Expert 5's BE PGA (GM 
model 616-A3) has significantly lower attenuation than the other 
models particularly at the larger magnitudes. This coupled with the 
site correction factor for rock increases the contribution from 
distant zones which have larger earthquakes and a relatively high 
rate of activity. 

(3) G-Expert 5 sets the BE value for the random uncertainty (standard 
deviation on the natural log of the PGA) to 0.7 compared to the range 
of values (0.35 - 0.55) selected by the other G-Experts. Relative to 
results obtained with a value of 0.55, this larger uncertainty (0.7) 
leads to an increase in the G-Expert 5's BEHC by about a factor of 2 
in probability of exceedance at lower (0.2g) g-values to over a 
factor of 3 at high g-values (0.9g). 

In summary we typically expect at rock sites that BEHC for G-Expert 5 for any 
S-Expert will be about a factor of 10-20 higher in probability of exceedance 
relative to the other BE GM models (factors (1) and (3) noted above) as 
illustrated in Fig. 2.3.1a. When the seismicity of the zones around the site 
are so low that there is very little contribution to the seismic hazard at a 
given site for a given S-Expert's input, then the low attenuation of G-Expert 
5's BE model, factor (2) becomes very important leading to the results shown 
in Fig. 2.3.1b. 



■34- 



Based on Figs. 2.3.1a and b we would expect that deleting G-Expert 5's GM 
model would have a significant impact on the hazard computed at many sites. 
To help understand the site to site variation observed in the sensitivity of 
the results to the inclusion/non-inclusion of G-Expert's 5 model, it is useful 
to note that the site can be placed into one of four categories: 

(1) Rock sites generally located in a region of low seismic activity (at 
least for some S-Experts) and located 200-600 km from a zone of very 
high activity with high upper magnitude cut off. Many sites fall 
into this category, see Vols. II-V. 

(2) Rock sites where the hazard is primarily from nearby zones. In 
particular see Vols. II and III. 

(3) Similar to (1) but a soil site. 

(4) Similar to (2) but a soil site. 

In Fig. 2.3.2a we illustrate a typical example for a category (1) site such as 
the Browns Ferry site where we reran the analysis only using G-Experts 1-4 GM 
models. We see from Fig. 2.3.2a that the 85th percentile CPHC is very 
sensitive to inclusion/non-inclusion of G-Expert 5's GM model. The median is 
less sensitive to the inclusion/non-inclusion of G-Expert 5's GM model and the 
15th percentile is even less sensitivity than the median. The AMHC is 
extremely sensitive to the inclusion/non-inclusion of G-Expert 5's GM model as 
can be seen from Fig. 2.3.2b. 

The sensitivity of the computed hazard to inclusion/non-inclusion of G-Expert 
5's GM model for sites that fall into category (2) is less than for category 
(1) sites as is illustrated in Fig. 2.3.3. The Limerick site used for the 
comparison is a typical rock site where the hazard is primarily from zones 
near the site. We see by comparing Fig. 2.3.2a to Fig. 2.3.3 that at the 
Limerick site the 85th percentile is much less sensitive to the G-Expert 5's 
GM model than at the Browns Ferry site. However, there is only a relatively 
small change in the sensitivity of the median and 15th percentile CPHCs to the 
inclusion/non-inclusion of G-Expert 5's GM model between the Browns Ferry site 
(category (1) type sites) and the Limerick (category (2) type sites). 

Unlike the differences in the sensitivity of the computed hazard to the 
inclusion/non-inclusion of G-Expert 5's model between category (1) and (2) 
type of sites, there is very little difference in the sensitivity of the CPHCs 
to the inclusion/non-inclusion of G-Expert 5's GM model between category (3) 
and (4) type of sites, as can be seen by comparing Figs. 2.3.4 to 2.3.5. In 
Fig. 2.3.4 the River Bend site is a typical example of a category (3) site and 
the Salem site used for Fig. 2.3.5 is a typical example of a category (4) 
site. We see by comparing Fig. 2.3.4 to Fig. 2.3.5 that there is little 
difference between the computed 85th percentile CPHC between including or not 
including G-Expert 5's GM model. The sensitivity of the median CPHC to G- 
expert 5's GM model is about the same for category (3) and (4) type of sites 
and only slightly less than for category (2) type of sites. 

-35- 



One of the main reasons for the impact of G-Expert 5's GM model on the median 
for all cases is reason (2) (high value of the random uncertainty) in the list 
of three reasons given earlier for why G-Expert 5's model tended to be higher 
for many S-Experts input as compared to the other G-Experts' models. 

The sensitivity of the CPUHS to G-Expert 5's GM model is approximately the 
same as illustrated for PGA, however, it "looks" different because in the 
comparisons between the estimated CPHCs the parameter of interest was the 
probability of exceedance whereas for the CPUHS, comparisons are made relative 
to actual spectral velocity. This is illustrated in Fig. 2.3.6 where we 
compare the 10,000 year return period CPUHS for the Browns Ferry site between 
the case when G-Expert 5's GM model is included and the case when it is not 
included. 

One important aspect of this comparison, in Fig. 2.3.6, is that the spectral 
shape is not changed by including or not including G-Expert 5's ground motion 
model. This point is repeated in Section 3.3 in the regional spectral 
comparisons. 

Because of the significance of the model selected by G-Expert 5 it might be 
worthwhile to review some of the good and bad aspects of the model relative to 
other models selected by the G-Experts. These comments are distilled from 
points made in our Ground Motion Questionnaires (i.e., questionnaires Q4, Q6 
and QIO which are given in Vol. 7). First, it should be noted that the model 
selected by G-Expert 5 is a data-based model. The model is made up by 
selecting a smoothed attenuation of intensity as a function of epicentral 
intensity and distance. G-Expert 5 selected the attenuation model developed 
by Gupta and Nuttli (1976) and modified as suggested in Q4 by a reduction in 
0.5 intensity units to account for the fact that Gupta and Nuttli 's worked 
with isoseismals rather than the median distances. The selected attenuation 
model was labeled model A3 in QIO. This attenuation of intensity model was 
based on several larger EUS earthquakes including the 1811-12 New Madrid 
series. No estimate of the "error of the fit of the model is given by Gupta 
and Nuttli (1976). However, typical fits of the attenuation of sensitivity 
generally yield a random error of over one intensity unit. 

In order to be used in our analysis, it is necessary to convert site intensity 
to a ground motion estimate. G-Expert 5 agreed that all the methods of doing 
this are flawed, however, the "best" approach, given the data available and 
the known differences between the EUS and WUS attenuation of ground motion, is 
to convert I^ to ground motion directly. There are many possible ways of 
doing this conversion as indicated in Q4 and QIO of Vol. 7. In all cases one 
lacks data to account for the significant differences in attenuation between 
regions for which there is sufficient data to construct a relation between I^ 
and ground motion for a range of epicentral intensities. In addition there is 
insufficient data to develop refined models to account for the difference 
between magnitude scale used in the various regions. G-Expert 5 selected the 
relation G-16 of QIO between intensity and PGA (Trifunac 1976) and the model 
TL-RS (Trifunac and Lee (1985)) between intensity and spectral values. 



-36- 



The model selected by G-Expert 5 is entirely data based which is more than can 
be said for the other models except for the lightly weighted model Comb-IA. 
In addition the model selected by G-Expert 5 and the lightly weighted Comb-IA 
are the only intensity based models selected. The argument for intensity 
based models is that they are the only direct ground motion data we have from 
the larger EUS earthquakes. The argument against intensity based models Are 
that: (1) they have poor correlation, i.e., one needs a large value for the 
error term, e.g., the value selected by G-Expert 5 is consistent with data, 
and if anything smaller, and (2) one does not have the data to develop the 
proper relation between distance ground motion and site intensity to 
substitute into the attenuation of intensity with distance relation. 

The fact that results obtained by using the ground motion model of G-Expert 5 
(labeled G16-A3) appear in many ways different, in general higher, than when 
using the other models motivated a careful reanalysis of this model. Our 
first step in quality control was to perform extensive auditing of all the 
questionnaire responses and to perform numerous sensitivity analyses to 
analyze the behavior of G16-A3 with respect to the seismic hazard, to ensure 
that everything we would observe was consistent with our understanding of the 
data. 

Then, in addition to the formal feedback questionnaires and meetings organized 
in this project (Q4, Q7 and QIO) , we had an extra one-on-one feedback meeting 
with G-Expert 5. In that meeting, we described again the versions of the 
models used in our analysis and studied together the results of the 
sensitivity analyses. We emphasized how the sites corrections were used as 
well as the type of seismicity and zonation models used in the analysis. 

This quality control confirmed that the models attributed to G-Expert 5 and 
the way we used them was consistent with Expert's 5 understanding of the 
problem and of the project's limitations. Since G16-A3 carries approximately 
20 percent of the weight of the ground motion models, we reviewed the various 
elements of this specific model which make it desirable or undesirable for use 
in a seismic hazard analysis for the E.U.S. 

In essence, there are three parts to the derivation of G16-A3. They are: 

1. The model of attenuation 

2. The scaling of earthquake size 

3. The correlation between site intensity and ground motion acceleration 
or velocity 

In addition (item 4), the correlation between magnitude and intensity. 
Although not really a part of the model, a relationship is used in our 
analysis. When the seismicity parameter are expressed in term of magnitude, 
the relationship between magnitude and intensity is used to convert intensity 
into magnitude, and vice-versa when the seismicity is in intensity and the 
ground motion model is in magnitude. Items 1 and 2 are relying entirely on 
E.U.S. intensity data since it is modeled by the Gupta-Nuttli relationship. 

-37- 



Item 3 relies mostly on strong motion data from the Western U.S. As such, one 
might question its applicability to the E.U.S. To our knowledge, however, 
there is no definitive work showing that the use of Western U.S. strong motion 
data to correlate site intensity with PGA/or PGV would not apply to the E.U.S. 

Some of the sites in the E.U.S. are located on rock (approximately 47% of 
them), some are located on deep soil (approximately 17% of them) and the 
remaining ones (approximately 36% of them) are located on shallow soil. More 
research is needed at the present time to clarify whether there is significant 
difference, for each site condition considered, in the relationship of PGA (or 
PGA) versus I^, between the Eastern U.S. and the Western U.S., for the same 
soil conditions and for the same site intensity (I5). 

The additional item mentioned above (item 4) is a relationship between 
magnitude and epicentral distance which was developed with Eastern U.S. data. 

Given the paucity of ground motion data from even relatively small earthquakes 
(up to mu=5) in the EUS and the total lack of measured data from large EUS 
earthquakes, it is in our view necessary to include intensity based models in 
the analysis. However, it must also be recognized that in an analysis such as 
we are performing, certain combinations of assumptions will lead to estimates 
that are true outliers. Ideally in such cases one would review both the 
seismicity assumptions, ground motion model assumptions and/or the weights 
selected to see if there isn't a better set of seismicity and ground motion 
models and weights that should be used. Clearly, in our case where we have 
one set of experts providing the seismicity input and a different set 
providing the ground motion models, it is difficult to do this reconsideration 
of the input. This in our opinion makes the AMHC a relatively poor choice to 
use to compare the hazard between sites because it is more sensitive to 
outliers than other estimators, such as the median, or other percentiles. On 
the other hand, given the spread between expert opinion observed in this study 
the variation in the median CPHC is relatively small between the case when S- 
Expert 5 is included or not included. This in our view makes the median 
estimate a very desirable estimator, given its relatively high stability. 
Thus we would recommend making the needed assessments and comparisons between 
sites relative to the median CPHCs and median CPUHS with all of the S- and G- 
Experts included. However, any assessment must account for the large 
uncertainty in the seismic hazard that exists. 



-38- 



LOWER MAGNITUDE OF INTEGRATION = 5. 



-1 
10 



-2 
10 

< 

UJ 

>- 

I.I o 

Q^ 10 

UJ 

o 

z 
< 
o 

UJ A 

UJ -4 

o 10 

X 

Ul 

u. 

o 

H -5 

-J 10 

CD 
< 

m 
o 
q: 

a. 



BEST ESTIMATES FOR SEISMIC EXPERT 3 
HAZARD CURVES BY ATTENUATION EXPERT 



-6 
10 



-7 
10 




o 

u, ACCELERATION CM/SEC* ♦ 2 



BROWNS FERRY 



Figure 2.3.1a Comparison between the BEHCs per G-Expert for S-Expert 3's 
input for the Browns Ferry site. 



-39- 



LOWER MAGNITUDE OF INTEGRATION = 5. 

BEST ESTIMATES FOR SEISMIC EXPERT 4 
HAZARD CURVES BY ATTENUATION EXPERT 



10 



-2 
10 

< 

Ul 

>- 

LrJ •J 
Q. 10 



< 

LJ ^ 

o 10 



CO 

< 

m 
o 
q: 



-5 
10 



-6 
10 



-7 
10 




ACCELERATION CM/SEC* '2 

BROWNS FERRY 



Figure 2.3.1b Comparison between BEHCs per G-Expert for S-Expert 4's input 
for the Browns Ferry site. 



-40- 



LOWER MAGNITUDE OF INTEGRATION IS 5.0 

PERCENTILES = 15.. 50. AND 85. 



10 





-2 




10 


a: 




< 




LiJ 




>- 






-3 


Q. 

> — •' 


10 


LJ 




(J 




z 




< 




o 




L4J 


-4 





10 



ll -5 
-J 10 

m 
< 

m 
O 
or 
a. 

-6 
10 



10 




ACCELERATION CM/SEC»*2 

BROWNS FERRY 



Figure 2.3.2a Comparison between the CPHCs when all 5 G-Experts are used and 
when G-Experts is not included for the Browns Ferry site. 



-41- 



-1 

10 



-7 
10 



I I 

B-BEST.A-ARITHMETIC 



• Only G-Expert 
- Al ] 5 G-Exper 




]-h 



o 

+ 



to -* IT) U3 

ACCELERATION CM/5EC»*2 



BROWNS FERRY 



Fieure 2 3 2b Comparison between the AMHCs and BEHCs when all 5 G-Experts are 
Figure 2.3.2b Comp ^^^ ^^^^ ^_^^^^^^ ^ .^ ^^^ .^^^^^^^ ^^^ ^^^ 3^^^„3 ,,,,y 

site. 



-42- 



PERCENTILES = 15.. 50. AND 85. 



-1 
10 



-2 
10 

< 

UJ 

>- 

£L 10 



U 

z 
< 

^ -I 

Ul 

o 10 

X 



o 

>- 



CD 
< 

CD 
O 

or 



10 



-6 
10 



-7 
10 



Only G-Experts \-^ 
Al 1 5 G-Experts , 




O 

+ 



ro "* in (O 

ACCELERATION CM/SEC»»2 

LIMERICK 



Figure 2.3-3 



Comparison between the CPHCs when all the G-Experts are used 
and when G-Expert 5 is not included for the Limerick site. 



-43- 



PERCENTILES = 15.. 50. AND 85. 



< 
ui 

>- 

UI 



z 
< 



O 
X 



< 

CD 
O 

a: 
a. 



-1 
10 



-2 
10 



-3 

10 



-4 
10 



-5 
10 



-6 
10 



-7 
10 



n r 



Only G-Experts 1-^ 
Al 1 5 G-Experts 




ACCELERATION CM/SEC* »2 

RIVER BEND 



Figure 2.3. i< 



wherGlxnP^f r'" "'l^'" "'"" "'' "' '^" G-Experts are used and 
when G-Expert 5 is not included for the River Bend site. 



-44- 



PERCENTILES = 15.. 50. AND 85. 



a. 

< 

>- 

Q£ 

UI 
Ol 



O 

z 
< 



10 



10 



-3 
10 



o 10 



O 



CD 

< 
CD 
O 

or 
a. 



10 



-6 
10 



-7 
10 



o 

+ 

UJ 



-♦ Only G-Experts 
~ Al 1 5 G-Experts 




\-h 



ro ■* in »D 

ACCELERATION CM/SEC* » 2 

SALEM 



Figure 2.3.5 



Comparison between the CPHCs when all of the G-Experts are used 
and when G-Expert 5 is not included for the Salem site. 



-45- 



LOWER MAGNITUDE OF INTEGRATION IS 5.0 
10000. -YEAR RETURN PERIOD CONSTANT PERCENTILE SPECTRA FOR 
PERCENTILES = 15.. 50. AND 85. 



10 



o 

ui 



o 
o 



2 
10 Ir 



10 



10 



-1 
10 



CM 

I O 



■n- 



-• Only G-Experts 1-^ 



Al 1 5 G-Experts 




I 



CN4 lo ■* in ir>r->j3aT> 
To 



PERIOD (SEC)'^2 

BROWNS FERRY 



""o 



Figure 2.3-6 



Comparison between the 10,000 year return period CPUHS when all 
of the G-Experts are used and when G-Expert 5 is not included 
for the Browns Ferry site. 



-46- 



2.4 Apparent Disconnect Between the PGA Hazard and the Spectral Hazard 

Based on a number of studies, e.g., Newmark and Hall (1978), it has been the 
custom to assume that at approximately 33Hz the spectral amplification of the 
PGA is unity. In Fig. 2.4.1 we have converted the CPHCs at the appropriate 
probability of exceedance to spectra values and plotted them on the 10,000 
return period CPUHS for the Limerick site. It is evident from Fig. 2.4.1 that 
there is a "disconnect" between the high frequency end of the CPUHS, the PGA 
and the assumed spectral amplification factor at 33Hz. The apparent 
disconnect arises because of the EUS random vibration models, which are 
heavily weighted by our G-Experts, have much higher amplification factors at 
short periods than typical western U.S. earthquakes. This is illustrated in 
Fig. 2.4.2. In Fig. 2.4.2 we compare the spectral model based on a Newmark- 
Hall model, i.e., standard western U.S. spectral amplification factors, 
developed using a random vibration PGA and velocity models {see Vol. 1 models 
RV-5A and RV-5V) and the resultant spectral model based on the same 
assumptions (model RV-5RS) used to develop the PGA and velocity models used 
with the Newmark-Hall amplification factors. We see that the resultant 
spectral shapes are significantly different. The random vibration spectral 
model has significantly more short period content than the Newmark-Hall 
spectral model and significantly less longer period energy than the Newmark- 
Hall model. This difference between the RV-spectral models and either 
Newmark-Hall type models or the model selected by G-Expert 5 lead to the 
results shown in Fig. 2.4.1 where the CPUHS at 25Hz are high relative to the 
PGA hazard curve. Additionally, the long period difference leads to the 
increased uncertainty at 1 sec as compared to 0.04 sees in our CPUHS typically 
plotted in Vols. II-V. See QIO in Vol. VII for additional discussion of the 
difference between the RV models and "typical" western U.S. spectral models. 

It should be noted that this "disconnect" may have important implications for 
the stiff components of nuclear power plants. For such equipment (natural 
period shorter than 0.05 sec.) it has been customary to assume that the 
spectral acceleration they are subject to is equal to the PGA. Clearly, as 
can be seen from Figs. 2.4.1 and 2.4.2 that this is a poor assumption in the 
EUS. The EUS RV-spectral models amplify the PGA for periods longer than 
0.1 sec. Typically, the PGA should be converted to spectral acceleration and 
plotted at 0.01s. The spectrum between 0.01 sec. and 0.04 sec. (last point 
computed) can be approximated by connecting the 0.01 spectral value estimated 
by converting the PGA hazard to a spectral velocity with unity amplification 
factor with the 0.04 value of the CPUHS plotted on the various figures by a 
straight line. This is illustrated in Fig. 2.4.3. 



-47- 



E.U.S SEISMIC HAZARD CHARACTERIZATION 
LOWER MAGNITUDE OF INTEGRATION IS 5.0 

10000. -YEAR RETURN PERIOD CONSTANT PERCENTILE SPECTRA FOR 

PERCENTILES = 15.. 50. AND 85. 



o 

(/) 
\ 

o 



o 
o 




PERIOD (SEC) o 

LIMERICK 



Figure 2.4.1 



Comparison between the 10,000 year return period CPUHS for the 
Limerick site and the spectral value estimated from the CPHCs 
using the Newmark-Hall amplification of 1.0 at 0.03 sec. to 
convert acceleration to relative velocity. 



-48- 



10 



o 

ui 
to 



2 
o 



o 
o 



UI 

> 



10 



10 



10 



-1 
10 




to ■<* IT) i£)r-~oar> 



eg rO •* lOCOrvcOT) 



fO •>* m i£>rv0O7> 



I o 



' ° PERIOD (SEC) 2 



Figure 2.M.2 



Comparison between the spectra obtained using the random 
vibration spectral model RV-5RS at an epicentral distance of 
15km for magnitudes 5. 6 and 7 and the spectra obtained using 
the Newmark-Hall median spectral amplifications applied to the 
random vibration acceleration model RV-5A and random vibration 
velocity model RV-5V. The models RV-5A and 5V are consistent 
with the spectral model RV-5RS. 



-49- 



E.U.S SEISMIC HAZARD CHARACTERIZATION 
LOWER MAGNITUDE OF INTEGRATION IS 5.0 

10000. -YEAR RETURN PERIOD CONSTANT PERCENTILE SPECTRA FOR 

PERCENTILES = 15.. 50. AND 85. 



10 



o 
t/» 



u 
o 



10 



-1 

10 



From CPHCs 




1 



CM 

I O 



To 



mJ. 



PERIOD (SEC) °o 

LIMERICK 



o 



Figure 2.4.3 



Illustration of how the spectral values for periods shorter 
than 0.04S can be estimated. The appropriate PGA value is read 
from off the CPHCs at the appropriate return period, converted 
to spectral velocity and plotted at 0.01s. Then a straight 
line, as shown, is used to connect the PGA value to the last 
computer CPUHS value at O.O^s. 



•50- 



3. COMPARISON BETWEEN SITES AND REGIONAL OBSERVATIONS 

3.1 General Comparisons Between Sites 

In Fig. 3.1.1a we plot the median CPHC for all 69 sites included in this 
study In Figs. 3.1.1b-e we plot the median hazard curves for each of the 
batches individually. The plot symbols used in Figs. 3.1.1b-e correspond to 
the identifications given in Tables l.la-d respectively. We see that there is 
a relatively wide spread between the site with the lowest and highest median 
CPHC. In Fig. 3.1.1a the two sites with highest median CPHC are the Seabrook 
and Pilgrim sites. However, we have pointed out a number of times that 
because the uncertainty is large the use of other estimators could lead to a 
different ordering of the sites. This point is illustrated in Fig. 3.1.2a 
and b for which the key to the identity of the sites is given in Tables 3.1.1 
and 3.1.2. In Fig. 3.1.2a, b we plot for each site the median (plot symbol 
M), the best estimate (B), the arithmetic mean (A) and the 15th and 85th 
percentiles (*) annual probabilities of exceeding 0.2g PGA. In Fig. 3.1.2a 
the sites are ordered by Volume; e.g., the sites in Vol. II are 1-19, m Vol. 
Ill they are 20-36, in Vol. IV they are 37-52 and in Vol. V they are 53-69. 
In Fig. 3.1.2b the sites have been ordered by the median hazard at 0.2g. This 
ordering, as well as the original ordering, is given in Tables 3.1.1 and 
3 12 We see from Fig. 3.1.2b that a very different ordering of the sites 
wouldVesult if we used say the BEHC or the AMHC. Clearly, the AMHC shows the 
largest variation of any of the estimators. We have pointed this out a number 
of times in Vols. II-V in our discussions of the various sites; e.g., see Vol. 
V Section 2.16 for the Waterford site (#68 in the ordering in Fig. 3.1.2a). 

We see from either Fig. 3.1.2 a or b that at the 0.2g level there is 
approximately 2 orders of magnitude spread between the 15th and 85th 
percentile CPHCs. In fact the largest spread between the 15th and 85th 
percentile CPHCs at 0.2g occurs at the Callaway site (#54 in Fig 3.1.2a and 
#46 in Fig. 3.1.2b) where the spread is 2.5 orders of magnitude. The smallest 
spread between the 15th and 85th CPHCs at 0.2g occurs at the Zion site (#52 in 
Fig. 3.1.2a and #32 in Fig. 3.1.2b) where the spread is 1.8 orders of 
magnitude. It is somewhat interesting to note that the site with the largest 
and smallest spread between the 15th and 85th CPHCs at 0.2g are both in the 
Central Stable region. We also see from Fig. 3.1.2, on the average, that the 
spread between the median CPHCs for all sites at 0.2g is about 1.4 orders of 
magnitude. 

We see from Figs. 3.1.1a-e that at higher g-levels a number of the median 
hazard curves cross, e.g., at low g values Seabrook (symbol E in Fig. 3.1.1b) 
has a higher median hazard for PGA than the Pilgrim site (symbol C in 
Fig, 3.1.1b), but at higher g-values the Pilgrim site has a higher median 
hazard for PGA. We also see from Figs. 3.1.1a-e that at higher g levels, the 
spread between the median CPHCs becomes larger. However, the uncertainty also 
grows. For example, Fig. 3.1.3 (for which the key to site identity is given 
in Table 3.1.3), shows that at the 0.6g level there is a 1.6 order of 
magnitude spread between the median hazard values from the site with the 
highest median hazard to the site with the lowest median hazard. The largest 

-51- 



spread between the 15th and 85th percentile CPHCs at 0.6g is 3.4 orders of 
magnitude (for the Grand Gulf site, #63 in Fig. 3.1.3) and the smallest spread 
rirfh. ?H 15th and 8 th percentile CPHCs at 0.6g is 2.1 orders of magnitude 
(for the Three Mile Island site, site #15 in Fig. 3.1.3). Thus the 
uncertainty at 0.6g has grown at a slightly larger rate than the variation 
between the median CPHCs. It is also important to note from Fig. 3.1.1a that 
at say the 10 probability of exceedance level, the median PGA for the sites' 
only varies by a factor of 4 whereas (as previously noted) at the 2g level 
the median probability of exceedance varies by a factor of 26 Thus the 
somewhat large variations in probability of exceedance that exist result in a 
much smaller uncertainty in the ground motion level. 



-52- 



-1 

10 



en 
< 


10 


UJ 




>- 




Q£ 




UJ 




Q. 




"^^ 


-3 


UJ 

o 


10 


z 




< 




Q 




UJ 




UJ 




o 




x 




UJ 


-4 


u. 


10 


o 




>- 




h- 




-J 




m 
< 


-5 


00 


10 


o 




on 




a. 






-6 




10 



-7 
10 




CM 
O 
+ 
UJ 



K) 



Tt in <i3 1^ 
ACCELERATION CM/SEC**2 



Figure 3.1.1a Comparison of the median CPHCs for all of the sites listed in 
Tables 1 .1a-d. 



-53- 



10 



< 


-I 
10 


UJ 




>- 




q: 




Ul 




CL 




"~~^ 


-3 


LJ 
O 


10 


z 




< 




Q 




UJ 




UJ 




O 




X 




UJ 


-4 


Li_ 


10 


o 




>- 




1— 




_I 




QD 
< 


-5 


CD 

O 


10 


a. 




CL 






-6 




10 



10 




T- CM (N 

O 
+ 
UJ 



Tt in to r^ 

ACCELERATION CM/SEC**2 



00 



en 



Figure 3.1.1b Comparison of the median CPHCs for the sites in Vol. II. The 
plot symbols used to identify the sites are given in Table 
1 .la. 



-54- 



-1 

10 



-2 
10 



-3 

10 



-4 
10 



< 



Q. 



o 
z 
< 

o 

UJ 
UJ 

o 

X 



o 

>- 



< ^ 

o 

0. 



-6 
10 



-7 
10 




T- (N 

O 

+ 



(N 



m 



rj- in (X) r~- 
ACCELERATION CM/SEC**2 



Figure 3.1.1c Comparison of the median CPHCs for the sites in Vol. III. The 
Figure i.i. P ^^^^ ^^^^ ^^ identify the sites are given in Table 



plot symbols 
1 .lb. 



-55- 



-1 

10 



,,-v 


-2 


< 


10 


UJ 




>- 




Q£ 




UJ 




Q. 




«>»<' 


-3 


o 


10 


z 




< 




o 




Lkl 




UJ 




o 




X 

UJ 


-4 


Li. 


10 


o 




>- 




h- 




_J 




CD 
< 


-5 


m 


10 


o 




IX. 

a. 






-6 




10 



-7 
10 




O 

+ 



CM 



-^ in i£) r\ 
ACCELERATION CM/SEC**2 



00 



(ji 



Figure 3.1. Id 



Comparison of the median CPHCs for the sites in Vol. IV. The 
plot symbols used to identify the sites are given in 
Table 1.1c. 



-56- 



10 



^-v 


-2 


< 


10 


LU 




>- 




q: 




UJ 




Q- 




' 


-3 


L±J 

o 


10 


z 




< 




Q 




LiJ 




UJ 




o 




X 




UJ 


-4 




in 


Lu 




O 




> 




1— 




_J 




CD 
< 


-5 


m 


10 


o 




q: 




Cl. 






-6 




10 



-7 
10 




O 

+ 



CN 



to 



-^t in to r--- 
ACCELERATION CM/SEC**2 



Figure 3. Lie Comparison of the median CPHCs for the sites in Vol. V. The 
plot symbols used to identify the sites are given in Table 
1 .1 .d. 



-57- 



TABLE 3.1.1 
KEY FOR SITES IN FIGURE 3.1.2a AND 3.1.2b 



(1) (2) (3) 

41 1 FITZPATRICK .5368E-04 

33 2 GINNA .5656E-04 

22 3 HADDAM NECK .1062E-03 

34 4 HOPE CREEK .6818E-04 
9 5 INDIAN POINT .1380E-03 

10 6 LIMERICK .1350E-03 

7 7 MAINE YANKEE .1838E-03 

23 8 MILLSTONE .1010E-03 
40 9 NINE MILE POINT .5374E-04 
43 10 OYSTER CREEK .5132E-04 

11 11 PEACH BOTTOM .1323E-03 

A li ^J!-25I" .3383E-03 

35 13 SALEM .6388E-04 
,1 14 SEABROOK .3406E-03 
31 15 SHOREHAM .8080E-04 
26 16 SUSQUEHANNA .9442E-04 
14 17 THREE MI. ISLAND .1219E-03 

19 18 VERMONT YANKEE .1132E-03 

5 19 YANKEE ROWE .2017E-03 

6 20 BELLEFONTE .2014E-03 

24 21 BROWNS FERRY .9930E-04 
13 22 BRUNSWICK .1245E-03 

47 23 CALVERT CLIFFS .4897E-04 

25 24 CATAWBA .9606E-04 

58 25 FARLEY .2932E-04 

55 26 HATCH .3767E-04 

20 27 MCGUIRE .n24E-03 
16 28 NORTH ANNA .1214E-03 

8 29 OCONEE .1818E-03 

29 30 ROBINSON 8762E-04 
,3 31 SEQUOYAH .2264E-03 
33 32 SHEARON HARRIS .6917E-04 

U il IHSSI'^ .1210E-03 

53 34 SURRY .4017E-04 

30 35 VOGTLE .8597E-04 
4 36 WATTS BAR .2151E-03 

18 37 BEAVER VALLY .n48E-03 

59 38 BIG ROCK POINT .2824E-04 

48 39 BRAIDWOOD .4660E-04 

44 40 BYRON 1 S 2 .5060E-04 
12 41 CLINTON .1309E-03 

36 42 COOK 1 8 2 .6330E-04 

37 43 DAVIS BESSE 1 .5826E-04 

45 44 DRESDEN 283 .4955E-04 

51 45 FERMI 2 .4437E-04 

56 46 KEWAUNEE .3434E-04 
28 47 LASALLE .8805E-04 
42 48 PALISADES 1 .5332E-04 
50 49 PERRY .4447E-04 

54 50 POINT BEACH .3936E-04 
H li §ySR CITIES .3244E-04 
32 52 ZION .7786E-04 

21 53 ARKANSAS .1123E-03 

46 54 CALLAWAY .4951 E-04 

H if ^2!:3St^S"E peak .1666E-04 

15 56 COOPER .1215E-03 

65 57 CRYSTAL RIVER .20nE-04 

63 58 DUANE ARNOLD .2126E-04 
27 59 FORT CALHOUN .9263E-04 
62 60 GRAND GULF .2175E-04 
39 61 LA CROSSE 5448E-04 

49 62 MONTICELLO 4510E-04 

52 63 PRAIRIE ISLAND 4189E-04 
61 64 RIVER BEND 2238E-04 
69 65 SOUTH TEXAS 1311E-04 
68 66 ST. LUCIE 1517E-04 

66 67 TURKEY POINT .'l934E-04 

64 68 WATERFORD 2045E-04 
60 69 WOLF CREEK 2807E-04 



Column (1) is for the site numbers for Fig. 3.1.2b 

Column (2) is for the site numbers in Fig. 3.1.2a (i.e., order in Vol. I) 

Column (3) is the median hazard value of 0.2g 



-58- 



TABLE 3.1.2 

KEY FOR SITES IN FIGURE 3.1.2a AND 3.1.2b 

AND MEDIAN VALUES AT 0.2g 



(1) (2) (3) 



1 


14 


SEABROOK 


.3406E-03 


2 


12 


PILGRIM 


.3383E-03 


3 


31 


SEQUOYAH 


.2264E-03 


4 


36 


WATTS BAR 


.21 51E-03 


5 


19 


YANKEE ROWE 


.2017E-03 


6 


20 


BELLEFONTE 


.2014E-03 


7 


7 


MAINE YANKEE 


.1838E-03 


8 


29 


OCONEE 


.1818E-03 


9 


5 


INDIAN POINT 


.1380E-03 


10 


6 


LIMERICK 


.1350E-03 


11 


11 


PEACH BOTTOM 


.1323E-03 


12 


41 


CLINTON 


.1509E-03 


13 


22 


BRUNSWICK 


.1245E-03 


14 


17 


THREE MI. ISLAND 


.1219E-03 


15 


56 


COOPER 


.121 5E-03 


16 


28 


NORTH ANNA 


.1214E-03 


17 


33 


SUMMER 


.1210E-03 


18 


37 


BEAVER VALLY 


.1148E-03 


19 


18 


VERMONT YANKEE 


.1132E-03 


20 


27 


MCGUIRE 


.1124E-03 


21 


53 


ARKANSAS 


.1123E-03 


22 


3 


HADDAM NECK 


.1062E-03 


23 


8 


MILLSTONE 


.1010E-03 


24 


21 


BROWNS FERRY 


.9930E-04 


25 


24 


CATAWBA 


.9606E-04 


26 


16 


SUSQUEHANNA 


.9442E-04 


27 


59 


FORT CALHOUN 


.9263E-04 


28 


47 


LASALLE 


.8805E-04 


29 


30 


ROBINSON 


.8762E-04 


30 


35 


VOGTLE 


.8597E-04 


31 


15 


SHOREHAM 


.8080E-04 


32 


52 


ZION 


.7786E-04 


33 


32 


SHEARON HARRIS 


.6917E-04 


34 


4 


HOPE CREEK 


.6818E-04 


35 


13 


SALEM 


.6388E-04 


36 


42 


COOK 1 S 2 


.6330E-04 


37 


43 


DAVIS BESSE 1 


.5826E-04 


38 


2 


GINNA 


.56 56E-04 


39 


61 


LA CROSSE 


.5448E-04 


40 


9 


NINE MILE POINT 


.5374E-04 


41 


1 


FITZPATRICK 


.5368E-04 


42 


48 


PALISADES 1 


.5332E-04 


43 


10 


OYSTER CREEK 


.5132E-04 


44 


40 


BYRON 1 & 2 


.5060E-04 


45 


44 


DRESDEN 2&3 


.4955E-04 


46 


54 


CALLAWAY 


.4951E-04 


47 


23 


CALVERT CLIFFS 


.4897E-04 


48 


39 


BRAIDWOOD 


.4660E-04 


49 


62 


MONTICELLO 


.4510E-04 


50 


49 


PERRY 


.4447E-04 


51 


45 


FERMI 2 


.4437E-04 


52 


63 


PRAIRIE ISLAND 


.4189E-04 


53 


34 


SURRY 


.4017E-04 


54 


50 


POINT BEACH 


.3936E-04 


55 


26 


HATCH 


.3767E-04 


56 


46 


KEWAUNEE 


.3434E-04 


57 


51 


QUAD CITIES 


.3244E-04 


58 


25 


FARLEY 


.2932E-04 


59 


38 


BIG ROCK POINT 


.2824E-04 


60 


69 


WOLF CREEK 


.2807E-04 


61 


64 


RIVER BEND 


.2238E-04 


62 


60 


GRAND GULF 


.2175E-04 


63 


58 


DUANE ARNOLD 


.2126E-04 


64 


68 


WATERFORD 


.2045E-04 


65 


57 


CRYSTAL RIVER 


.201 1 E-04 


66 


67 


TURKEY POINT 


.1934E-04 


67 


55 


COMANCHE PEAK 


.1666E-04 


68 


66 


ST. LUCIE 


.1 517E-04 


69 


65 


SOUTH TEXAS 


.1311E-04 



Column (1) is for the site numbers for Fig. 3.1.2b 

Column (2) is for the site numbers in Fig. 3.1. 2d (i.e., order in Vol. I) 

Column (3) is the median hazard value of 0.2g 



-59- 



TABLE 3.1.3 

KEY FOR SITES IN FIGURE 3.1.3 AND MEDIAN 

HAZARD VALUES FOR 0.6g 



(1) (2) (3) 



1 


12 


PILGRIM 


.1844E-04 


2 


14 


SEABROOK 


.1269E-04 


3 


36 


WATTS BAR 


.7950E-05 


4 


19 


YANKEE ROWE 


.7693E-05 


5 


31 


SEQUOYAH 


.7658E-05 


6 


7 


MAINE YANKEE 


.7427E-05 


7 


20 


BELLEFONTE 


.6441E-05 


8 


22 


BRUNSWICK 


.6278E-05 


9 


56 


COOPER 


.5789E-05 


10 


37 


BEAVER VALLY 


.4779E-05 


n 


5 


INDIAN POINT 


.4684E-05 


12 


11 


PEACH BOTTOM 


.4677E-05 


13 


6 


LIMERICK 


.4451E-05 


14 


29 


OCONEE 


.4378E-05 


15 


17 


THREE MI. ISLAND 


.4221E-05 


16 


41 


CLINTON 


.4042E-05 


17 


59 


FORT CALHOUN 


.4005E-05 


18 


3 


HADDAM NECK 


.3986E-05 


19 


47 


LASALLE 


.3941E-05 


20 


53 


ARKANSAS 


.3797E-05 


21 


52 


ZION 


.3364E-05 


22 


8 


MILLSTONE 


.3345E-05 


23 


18 


VERMONT YANKEE 


.3204E-05 


24 


16 


SUSQUEHANNA 


.2877E-05 


25 


28 


NORTH ANNA 


.2849E-05 


26 


61 


LA CROSSE 


.2721E-05 


27 


42 


COOK 1 & 2 


.2714E-05 


28 


35 


VOGTLE 


.2655E-05 


29 


27 


MCGUIRE 


.2597E-05 


30 


33 


SUMMER 


.2592E-05 


31 


15 


SHOREHAM 


.2550E-05 


32 


30 


ROBINSON 


.2395E-05 


33 


48 


PALISADES 1 


.2327E-05 


34 


24 


CATAWBA 


.2285E-05 


35 


62 


MONTICELLO 


.1938E-05 


36 


63 


PRAIRIE ISLAND 


.1836E-05 


37 


13 


SALEM 


.1802E-05 


38 


43 


DAVIS BESSE 1 


.1802E-05 


39 


4 


HOPE CREEK 


.1757E-05 


40 


21 


BROWNS FERRY 


.1672E-05 


41 


50 


POINT BEACH 


.1658E-05 


42 


32 


SHEARON HARRIS 


.1634E-05 


43 


40 


BYRON 1 S 2 


.1383E-05 


44 


23 


CALVERT CLIFFS 


.1370E-05 


45 


10 


OYSTER CREEK 


.1367E-05 


46 


44 


DRESDEN 283 


.1362E-05 


47 


39 


BRAIDHOOD 


.1286E-05 


48 


46 


KEWAUNEE 


.1269E-05 


49 


45 


FERMI 2 


.1208E-05 


50 


2 


GINNA 


.1190E-05 


51 


49 


PERRY 


.1178E-05 


52 


38 


BIG ROCK POINT 


.1119E-05 


53 


26 


HATCH 


.1116E-05 


54 


34 


SURRY 


.1034E-05 


55 


9 


NINE MILE POINT 


.1002E-05 


56 


1 


FITZPATRICK 


.9798E-06 


57 


54 


CALLAWAY 


.8241E-06 


58 


25 


FARLEY 


.7470E-06 


59 


51 


QUAD CITIES 


.6566E-06 


60 


69 


WOLF CREEK 


.6046E-06 


61 


64 


RIVER BEND 


. 5831E-06 


62 


68 


WATERFORD 


. 5342E-06 


63 


60 


GRAND GULF 


. 5035E-06 


64 


58 


DUANE ARNOLD 


.4553E-06 


65 


57 


CRYSTAL RIVER 


.451 2E-06 


66 


67 


TURKEY POINT 


.3837E-06 


67 


55 


COMANCHE PEAK 


.3238E-06 


68 


66 


ST. LUCIE 


.3190E-06 


69 


65 


SOUTH TEXAS 


.2909E-06 



Column (1) is for the site numbers for Fig. 3.1.3 
Column (2) is for the site numbers as ordered in Vol. I 
Column (3) is the median hazard value of 0.6g 



-60- 



o 

z 
< 
o 

o 

X 

u 



m 
o 



< 

Z 
Z 
< 

O 

o 

Q 



-2 



AA 



-5 



h<yi 



I >^A 



MB 



M 



I 

HI 



-6 



-7 



H 



k 

I 

I! 



w 



u 



s^' 



m 



m 



A A 



.r, 



Ej^i 



hi 



A 



J^ 



'h^ 



H 



M 



hi 



M 



Ai 



M'1 



m 



ta 



'Hi, 



>^ A 
Af .A^JfjA 



M^ 



h. 



i 



?t|l 
hi 



M 



A*A A 



wk 



'^ 



B 



hi 



ll^ 



* 



'.W 



M 



o 

CM 



O 



O 



O 

in 



o 



o 
1^ 



SITE NUMBER 



Figure 3.1.2a 



Plot of the log of the annual probability of exceeding 0.2g 
for all the sites in Vols. II-V. The plot symbols are: 
A=arithmetic mean, M-median {50th percentile), {*) for the 
15th and 85th percentiles and B»best estimate. The sites are 
ordered by Volume. The key is given in Table 3.1.1 and cross 
referenced in Table 3.1.2. 



-61- 



o 
z 
< 

o 

LJ 

o 

X 

u 

u 
o 



m 
o 

0. 



z 
z 
< 

Li. 

o 

o 
o 



-2 



-3 



-4 



-5 



-7 



Aa a 



ii 



m 9 

I 



ma 



AA 



[JE^ 



^^YHlHI^y..^,,^^! 



EiB 



A At 



AA 



(JB 



* * 



^A 



'ffuii 



Ba 



m 



** 



fHHIUtiJiJ. 



^ A AAA 

Aa a 



;f^[j 



*A 



A 



f^U 



* ♦ A^ AaA 



A ♦ 



Ui'i' 



•Mo 



AA 



HIS 



o 

CM 



O 



O 



o 
in 



o 

ID 



o 



SITE NUMBER 



Figure 3.1.2b 



Same as Fig. 3.1.2a except the sites have been ordered by 
median probability of exceeding 0.2g. The ordering is given 
in Table 3.1.2 and cross referenced in Table 3.1.1. 



-62- 



o 

z 

UJ 

ui 

o 

X 



m 
o 



< 

3 



O 
O 



-3 



A A 



-4 



fin 



-5 






-6 



F% 



-7 



-8 



*aA 



^<^' 



>t^i, 



\n 



4< <^ 



aa ^ 

A. * * A 

A 

A 



T)h4 



A.f 



o 

CM 



t A 



"*mU 



u 



A A 

A^ 

A A 



•¥ A 



^'^w^im^ 



im 



o 



o 



tn 



A*A 



m 



o 
in 



t^ A ^AA 



'1^1. 



*\ ^ A 



[iffi 



l^¥t 



r>iH: 



o 
<x> 



o 



SITE NUMBER 



Figure 3.1.3 



Plot of the log of the annual probability of exceeding 0.6g 
for all of the sites in Vols. II-V. The sites have been 
ordered by the median probability of exceeding 0.6g. The plot 
symbols are: M=median, A=arithmetic mean, (*)=15th and 85th 
percentiles and B=best estimate. The ordering is given in 
Table 3.1.3. 



■63- 



3.2 Regional Comparisons (PGA) 

The median probability of exceedance of 0.2 g PGA for the sample of 69 sites 
across the EUS, shown in Fig. 3.1.2a, does not display any obvious 
characteristics that might correlate with the regional location of the 
sites. However, the grouping of sites used in Fig. 3.1.2a is not quite 
regional as explained in Vol. I, and in addition any such regional correlation 
could very well be obscured by the local site correction used in our analysis 
(see Section 2) . 

There are several natural groupings of the sites that might be interesting to 
compare. For example, it is interesting to compare the hazard at sites in New 
England to sites located either near the New Madrid region or the Charleston 
region. In making this comparison, we selected only rock sites. Because the 
Seabrook site has the highest median CPHC for PGA of the rock sites, we 
selected it and three other sites in New England relatively close to the 
Seabrook site. For the comparison between the Charleston and New Madrid 
areas, we considered it important to have the sites located approximately the 
same distance from either the New Madrid source region or the Charleston 
source region. In addition, our criterion of having rock sites for the 
comparison limited the number of sites. It should be noted, as can be seen by 
examining the various S-Experts' maps given in Appendix B, that there is a 
large variation in how the New Madrid and Charleston source zones are defined 
between the various S-Experts. Thus 1t 1s extremely difficult to define a 
meaningful distance metric to select the sites for comparison. We did not 
attempt to factor these complex considerations into our selection process in a 
rigorous way. Table 3.2.1 lists the sites we selected as best fitting our 
requirements and in addition to the New England sites, the three sites "near" 
Charleston and the three sites "near" New Madrid, we also selected three rock 
sites approximately mid-way between Charleston and the New Madrid region. 

In Fig. 3.2.1 we plotted the location of the sites listed in Table 3.2.1 
(indicated by the symbols X), the median PGA with a 10,000 year return period 
and the location of the Charleston (C) and New Madrid (NM) historic 
earthquakes. In Fig. 3.2.2 we plotted for each of the sites in Table 3.2.1 
the median (M), best estimate (B), arithmetic mean (A), and the 15th and 85th 
percentile value (*) annual probabilities of exceeding 0.2g. 

It is difficult to make statements about the relative difference between the 
New England region and sites "near" either the Charleston or New Madrid 
regions because if we examined locations closer to either of these two source 
zones, the seismic hazard would significantly increase over the values shown 
in Fig. 3.2.1 whereas it is unlikely that a site in New England would have a 
median seismic hazard level significantly higher than the level at the 
Seabrook site. 

We see from Fig. 3.2.2 that the hazard estimates for the three sites near the 
New Madrid region (site numbers 8, 9 and 10 in Fig. 3.2.2) are more uncertain 
than the estimates at the other sites. The make-up of the earthquakes 
contributing to the hazard are different for the three groupings of sites and 

-64- 



the range of size of possible earthquake motion at those sites is larger than 
at the other 7 sites considered in Fig. 3.2.2. The relative contribution of 
various earthquakes for each site is discussed in the appropriate section in 
Vols. II-V and plotted in Fig. 2.SN.4 for each site (where SN is the site 
number). Fig. 3.2.3 is a plot of the estimated 10,000 year return period PGA 
based on the BEHC estimator and including only earthquakes 6.5 and larger for 
each of the sites listed in Table 3.2.1. We see that the three sites near New 
Madrid are much higher than either for the sites near Charleston or between 
Charleston and New Madrid. The BE PGA hazard in New England is the smallest 
in this figure. From this we can infer that the hazard near New Madrid is 
primarily from very large earthquakes whereas for the Charleston area smaller 
local earthquakes are more important and thus level of hazard from large 
earthquakes is smaller for sites near Charleston than for sites around New 
Madrid region but larger than for sites in New England. To further illustrate 
this point and show that Figs. 3.2.4 agree with the above discussion we 
plotted the relative contribution to the BEHC for PGA only from earthquakes 
within 4 magnitude ranges for the Arkansas site (near New Madrid), the Catawba 
site (near Charleston), the Seabrook site (New England), and the Watts Bar 
site (between NM and C). The Figs. 3.2.4 were previously shown in 



in this study 
'tes are 
es by S. 



Figs. 3.2.4a, b, c, d in Vols. II-V. 

In Fig. 3.2.5 we plot the relative locations of all the sites in t 
and the median PGA with a 10,000 year return period. The rock sit 
denoted by R, the deep soil sites by D and the shallower soil site 

Fig. 3.2.5 indicates that, among those sites close to each other and therefore 
within similar seismotectonic environments, the 10,000 year return period PGA 
at shallow sites is higher than at rock sites, as expected from a prior 
discussion on the effect of correction for soil site conditions in 
Section 2.2. As indicated earlier, it is difficult to see any obvious trend 
in the effect of site correction, and if we only based our conclusions on 
Fig. 3.2.6 which shows the ratios of probabilities of exceedance of 0.2g 
between the shallow soil case and the rock case at the 12 sites, we would 
erroneously conclude that it is quite erratic and unpredictable. 

However, it is easier to understand the behavior of the results when the 
parameter of interest is the ground motion parameter itself (i.e., the PGA or 
PSRV) rather than the probabilities of exceedance. 

Let us first examine the ratios, r(p), of PGA between the shallow soil a^{p) 
to the rock case ap(p) for three different but fixed probabilities of 
exceedance (p) equal to 10""^, lO"'* and 10~^. r(p) is defined by: 

a,{p) 

The values of ^^{p) and ay,(p) reported in Table 3.2.3 (Table 3.1.1 of 
Vol. VIII) are taken from the Fig. 2.SN.4 of Vol. VIII for each of the 12 
sites in Table 3.2.3. 



-65- 



Recall that there are two types of site corrections being applied in this 
analysis (see Vol. I Section 3.7). 

One type of correction is the simple correction advocated by G-Expert 5, 
for which the median correction factor shallow soil/rock is approximately 
equal to 0.73 (see left side of Fig. 3.10 in Vol. I) regardless of the 
specific shallow site category. 

The other type of correction, advocated by the other 4 G-Experts, is the 
categorized correction for which the ratios (shallow/rock) depend on the 
soil category (see left sides of Figs. 3.12 and 3.13 of Vol. I) and are 
equal to: 



Sand-1/Rock 
Till-1/Rock 
Till-2/Rock 



r=1.65 
r-1.55 
r-1.38 



Thus if the only ground motion input used were that of G-Expert 5, we would 
expect the average correction factor r(p) to be always approximately 0.73. 

Furthermore, if G-Expert 5's input were not used, we would expect the average 
ratio shallow/rock to be 1.65 when the shallow soil is in category Sand-1, 
1.55 if it is in category Till-1, and 1.38 if it is in category Till-2. 

Since the results presented here used input from all the G-Experts in a 
proportion approximately of 1/5 weight for each of them, we would expect, on 
the average, the ratios r(p) to be equal to: 

(.73)(.20) + (1.65)(.80) = 1.47 for Sand-1/Rock 
(.73)(.20) + (1.55)(.80) = 1.39 for Till-1/Rock 
(.73)(.20) + (1.38)(.80) = 1.25 for Till-2/Rock 

Column (7) of Table 3.2.3 gives for each of the 12 sites of Table 1.1 the 
expected approximate ratio if G-Expert 5 were not used, column (6) shows the 
expected ratio if only G-Expert 5 were used, column (5) gives the expected 
approximate ratio if all G-Experts were weighted equally, and the next column 
(4) gives the average of the ratios shallow/rock given in columns (1), (2) and 
(3). Columns (1), (2) and (31 give the r(p) values for the probabilities of 
exceedance 10"^, 10"^ and 10 ^. 

Table 3.2.3 shows clearly that the effective correction factors (column (4)), 
which are obtained as an average of correction factors for three given 
probabilities of exceedance, are in general very close to the approximate 
values one would expect if the ground motion experts choices of correction 
were weighted equally (compare columns (4) and (5) in Table 3.2.3). 

The deviation from the value in column (5) is due to the complex interaction 
between ground motion models and seismicity zones, seismicity parameters and 
the fact that the correction factor is not deterministic but is defined by a 
probability distribution. Depending on all those factors the impact will be 
that the correction advocated by G-Expert 5 will have more or less weight, 
relative to the other 4 experts. For Oconee, the combination of the above 

-66- 



mentioned interactions leads to an impact of G-Expert 5 greater than the equal 

weight case. For the other sites, but Three Mile Island and North Anna, the 

effect is reversed and the opinion of G-Expert 5 appears to be more diluted 
than in the equal weight case. 

For Three Mile Island and North Anna neither group (i.e., with or without G- 
Expert 5) seems to dominate. 

The case of Arkansas, Callaway and Duane Arnold requires additional 
scrutiny. For those three sites. Table 3.2.3 shows that the effective 
amplification factors (column (4)) obtained in our simulation are close to the 
case when Expert 5's model is not used (compare column (4) with column (7)). 

This phenomenon seems extreme and can be explained as follows, (remembering 
that we are comparing median hazard curves for rock and for soil): 

For the rock case, the contribution to the hazard comes from distant large 
earthquakes. Figure 3.4 of Vol. I shows that in that range, G-Expert 5's 
ground motion model (number 3 on Fig. 3.4-Vol. I) is much higher than the 
rest of the models. Thus, the resultant median value is more 
representative of the other four ground motion models. 

For the shallow soil case, the large, distant earthquakes are also 

dominant, and G-Expert 5's model falls within the cluster of other models, 
thus, the median will be representative of all the models, and in 
particular again close to the median without Expert 5. 

The result is that the final ratio of PGA between shallow and rock cases for 
these three sites is close to the case when only the categorized correction is 
used (i.e., the correction recommended by all but G-Expert 5). 

Prior to drawing some conclusions, let us define the meaning of the term 
"correction" of the hazard curve. Let us assume that the hazard curve for a 
rock site is known, and one needs to have an estimate of the hazard at the 
same location but for a shallow soil condition. If one assumes the 
amplification from rock to soil to be a constant multiplicative value (say 
r^.) , then one would generate rigorously the soil hazard curve by taking each 
point of the rock hazard curve, say acceleration a|^ for a probability of 
exceedance h, and derive the corresponding point, ag, for the same probability 
of exceedance h, of the soil hazard curve such that 

^S "^ ^R * "^c ^^ constant h. 

Although this operation is correct for a constant r^ as indicated above, it 
would not be correct to perform it when a combination of correction types are 
used as in our study where the final effect is in between the two types of 
corrections as indicated in Table 3.2.3, and the relative weight of each type 
of correction depends both on the dominant zonation effects and on the 
dominant ground motion models. 



-67- 



However, Table 3.2.3 shows that constructing a soil hazard curve by first 
starting from our rock hazard curves and applying an average correction factor 
would lead to an estimated soil hazard curve close to the hazard curve 
estimated by our full method described in Vol. I and Section 2.2 of this 
volume. 

Table 3.2.3 shows that the error could be negligible in some cases, and at 
most, for the 12 sites considered here, the error would have been 13% (for 
Callaway). In all 12 cases but one (i.e., Oconee), the error would have been 
an underestimation (it would have been overestimated by approximately 3% at 
Oconee) . 

At the present time, we have not been able to derive any simple correlation 
between this effective amplification factor (column (4) of Table 3.2.3) and 
the zonation characteristics, location, soil conditions, or any other 
parameters specific to any given site, thus making impossible the rigorous 
transformation of our rock hazard curves into soil hazard curves in a simple 
way. 

And finally, one needs to caution the reader in extending the above 
conclusions to the probability of exceedance space. In spite of the 
remarkable stability of the correction factors shown in Table 3.2.3, 
Fig. 3.2.6 shows a quite different effect. Figure 3.2.6 shows the ratios as a 
function of both the average slopes of the hazard curves (soil and rock hazard 
curves) and the average amplification from rock to soil. If all sites 
exhibited exactly the same rock hazard curves, then Fig. 3.2.6 would be an 
exact representation of column (4) of Table 3.2.3. However, the slopes of 
those hazard curves are not exactly the same as 0.2g, thus Fig. 3.2.6 shows 
some deviation from column (4) of Table 3.2.3. The general shape of 
Fig. 3.2.6 is representative of the overall process and can be considered as 
some sort of a signature. 

If some elements of the zonation, seismicity or ground motion models were to 
be changed. Fig. 3.2.6 would change. In a sensitivity test, we removed ground 
motion Experts' 5 input and found that Fig. 3.2.6 was slightly changed but its 
general shape and level were preserved. 

We feel confident that the effects shown in Table 3.2.3 and Fig. 3.2.6 are 
realistic representations of the physical effects given our assumptions on the 
site correction methods, and not due to some unexpected parasitic software or 
numerical problems such as the choice of number of simulation, for we have 
performed numerous tests in previous studies to validate our operating 
parameters (Bernreuter et al., 1985). 



-68- 



Table 3.2.1 

Rock Sites Selected for the Comparison Between the Hazard for 
Sites Located in New England, near Charleston, near New Madrid, 
and Half-Way Between New Madrid and Charleston 



Location 


Site 
Name 

Plot 

# 


Median PGA (g) 

with a 
10,000 Year 
Return Period 


Vary 
Approximate 
Distance (km) 
NM C 


New England 


1 Maine Yankee 

2 Millstone 

3 Seabrook 

4 Vermont Yankee 


0.253 
0.197 
0.311 
0.205 


— 


_^ 


Near 
Charleston 


5 Catawba 

6 McGuire 

7 Oconee 


0.193 
0.204 
0.244 


— 


275 
300 
300 


New Madrid 


8 Arkansas 

9 Browns Ferry 

10 Callaway 


0.205 
0.195 
0.151 


300 
250 
300 


— 


Between 
NM & C 


11 Bellefonte 

12 Sequoyah 

13 Watts Bar 


0.256 
0.267 
0.267 


350 
400 
425 


600 
550 
550 



-69- 



CQ 



o 

iS 

I — I 

O -I 

§^ 

UJ o 



>= c 



< 
►-• o 

00 tt 



(/) 





Plot 
Symbol 


1—1 


CM 


ro 


^ 


LO 


<^ 


•tc 
O 

q: 

•r- 
O 

oo 

«/) 

o 

•r- 
■!-> 

to 


CPHCs 
at 0.3g 


• 


.— 1 
CM 


• 

CM 


to 


ro 


o 

• 


Average Ratio 
of EGAs at 

Probability 


CO 

to 

. 


O 
. 

r-4 


• 


ro 

• 
»-t 


00 

in 

r-t 


ro 

ir> 

. 
f— • 


Secondary 

Soil 

Category 


.—1 

1 
•o 

cz 
to 

00 


CM 

•r- 

t— 


t— t 

1 

T3 

C 

to 
oo 


r-H 
1 

■o 
c 
to 
oo 


<r-l 

1 

T3 

C 

«o 
oo 


t— 1 
1 

■o 
c 
to 
oo 


Results 
in 


Vol. II 
2.9 


Vol. II 
2.16 


Vol. II 
2.17 


Vol. Ill 
2.2 


Vol. Ill 
2.5 


Vol. Ill 
2.6 


Soil 

Category in 
Vols. II-V 


u 
o 
on 


u 
o 


o 


o 
o 
on 


o 
o 
on 


u 
o 
on 


Site 
Name 


Nine Mile 
Point 


to 

c 
c 

sz 

cr 
to 

13 
to 


Three Mile 
Island 


<u 
u. 

to 

c 

o 

CO 


to 

fO 

-t-> 
o 


B 

fO 
U- 




Section 
Number 


.—1 


CM 


m 


^ 


u-> 


kO 



-70- 



lU 



LL l-l 





00 to 




UJ 




ss 




H-J 




O _J 




»-oo 


CM lU 


H SOME S 
SOME ON 


• 3 

l-H 


t±j^ 


CO o 


P w 


:::o 








< 




00 




UJi£ 




h- o 




»-• o 




oott 




u. z 




oo 




1- 




00 





Plot 
Symbol 


t^ 


00 


cr. 


<c 


CO 


o 


.^ 

u 

o 

o 

«/) 
o 

Of 


CPHCs 
at 0.3g 


o 

• 


CM 


• 

ro 


LO 
CM 


Lf) 


ro 


Lf) 
O 1 

■f- o 

+-> r-H >, 

O) w o •>- 

«j to • «o 
I. aro J3 

0) 1 o 
> *♦- O 1- 
< O r-i Q. 


1-H 
t-H 


ro 

.—1 


ID 

• 


ir> 

f— 1 


cy> 

l-H 


O 
LD 

.— 1 


Secondary 

Soil 

Category 


1 

c 
00 


1 

{= 

«o 

OO 


t— 1 
1 

•o 

c 

oo 


r-l 
1 


«— 1 
1 

•o 

c 

oo 


1— t 
1 

•r— 


Results 
in 


Vol. Ill 
2.9 


Vol. Ill 
2.10 


Vol. Ill 
2.14 


Vol. V 
2.1 


Vol. V 
2.2 


Vol. V 
2.6 


Soil 

Category in 
Vols. II-V 


U 
O 

on 


u 
o 
a: 


a 
o 
on 


o 
o 


o 
o 


u 
o 


Site 
Name 


c 
c 
< 

o 


Oconee 


Summer 


C 


r— 
C_) 


Duane Arnold 




Section 
Number 


1 


00 


<y\ 


o 

I— I 


I— 1 

1 


CM 

r— 1 



ro 



U 
0) 

to 

(U 



CO 

o 



ro 

o; 

+-> 

«♦- 
o 

V) 

c 
o 

•r- 
-M 
ID 



«o 

U 



0) 



c 
o 



(O 



■o 



o 



0) 

o 



•71- 



TABLE 3.2.3 

RATIOS OF PGA VALUES BETWEEN SHALLOW AND ROCK CONDITIONS 
FOR FIXED VALUES OF THE HAZARD 





Soil 
Category 


Ratio Shallow/Rock 


All 
Equal 
Weight 


Only 
G5* 




Site 


10"^ 


10-4 


10"5 


Avg. 


W/0 
65** 




(1) 


it) 


(3) 


(4! 


(5) 


(6) 


in 


1 Nine Mile Point 


Sand-1 


1.57 


1.58 


1.59 


1.58 


1.47 


0.73 


1.65 


2 Susquehanna 


Till-2 


1.30 


1.30 


1.30 


1.30 


1.25 


0.73 


1.38 


3 Three Mile Island 


Sand-1 


1.50 


1.47 


1.44 


1.47 


1.47 


0.73 


1.65 


4 Browns Ferry 


Sand-1 


1.56 


1.66 


1.68 


1.63 


1.47 


0.73 


1.65 


5 Catawba 


Sand-1 


1.59 


1.58 


1.55 


1.57 


1.47 


0.73 


1.65 


6 Farley 


Sand-1 


N/A 


1.56 


1.49 


1.53 


1.47 


0.73 


1.65 


7 North Anna 


Sand-1 


1.51 


1.50 


1.51 


1.51 


1.47 


0.73 


1.65 


8 Oconee 


Sand-1 


1.37 


1.44 


1.47 


1.43 


1.47 


0.73 


1.65 


9 Summer 


Sand-1 


1.47 


1.62 


1.61 


1.57 


1.47 


0.73 


1.65 


10 Arkansas 


Till-1 


1.51 


1.50 


1.50 


1.50 


1.39 


0.73 


1.55 


11 Callaway 


Sand-1 


1.65 


1.70 


1.72 


1.69 


1.47 


0.73 


1.65 


12 Duane Arnold 


Till-1 


N/A 


1.50 


1.50 


1.50 


1.39 


0.73 


1.55 



* Ratio of PGA shallow/rock given by G-Expert 5 only 

** Ratio of PGA shallow/rock given by G-Experts 1,2,3 and 4 only 



-72- 



45 


e^^'^^ 

f 


r^?v 


x.21 X.i2 


40 


X.15 

NM 




f 


35 


X.21 


X.27 ^ ,, 
X.20<-26 X.25 


^ 


30 


- 


'y\ 


- 


25 




^^ 


1 1 



in 


o 


IT) 


O 


in 


o 


1 


<7) 

1 


00 
1 


00 

1 


1 


1 



Figure 3.2.1 



The location of the sites listed in Table 3.2.1 is shown by 
the symbol X relative to the historic New Madrid (NM) and 
Charleston (C) earthquakes. The 10,000 year return period PGA 
median g-values are also shown. 



-73- 



o 



< 






CD 
O 

q: 



< 
=> 

2 
Z 
< 



O 

o 



-1 



-2 



-3 



-4 



-5 



A A 



A A 



M i; 



? ^ $ 



W 



I! i) (I li 



M 



I! I! 

►!< ^< M 



SITE NUMBER 



Figure 3.2.2 



A plot of the log of the annual probability of exceeding 0.2g 
for each of the sites listed in Table 3.2.1. The site number 
is given in Table 3.2.1. The plot symbols are: M=median, 
A=arithmetic mean, (*)=15th and 85th percentiles and B=best 
estimate. 



-74- 



500i 



# # # 



4« 
iH 
Ul 
\ 

E 
U 
I 

o 



i. 

(U 
01 

u 
u 
cc 



400- 



300- 



200- 



100- 



# 



# 



# # 



0^ 



Mill V.Y. nc6. B.F. Call. Seq. 
riaineY.Seab. Cat. Ocnee Rrk. Belf. W.B. 



Figure 3.2.3 



A plot of the estimated 10,000 year return period PGA only 
including large earthquakes of magnitude 6.5 and greater based 
on the BEHCs for each site listed in Table 3.2.1. 



-75- 



HAZARD CURVES USING ALL EXPERTS 



10 



-2 
10 

< 
>- 

I.I *J 

Ou 10 



Ul 

u 

z 

o 10 

X 



L -5 

_j 10 

m 

< 
m 
o 
on 
a. 



-6 

10 



-7 
10 



I I I I I 

CURVE 1 3.75<MB<5.0 

CURVE 2 5.0<MB<5.75 

CURVE 3 5.75<MB<6.5 

CURVE 4 6.5<MB 




-- (N 

o 



■* m »D t^ 
ACCELERATION CM/SEC* *2 

ARKANSAS 



Figure 3.2.4a 



BEHCs which include only the contribution to the PGA hazard 
from earthquakes within the indicated magnitude range for the 
Arkansas site (near the New Madrid area). 



-76- 



CONTRIBUTION TO THE HAZARD FOR PGA 

FROM THE EARTHOUKES IN 4 MAGNITUDE RANGES 



< 
ui 

>- 

UI 



u 

z 

2 



o 

X 



o 

>- 



< 

m 
o 
q: 

a. 



-1 
10 



-2 
10 



-3 

10 



-4 
10 



-5 
10 



-6 
10 



-7 
10 



HAZARD CURVES USING ALL EXPERTS 




-t irt >£> r^ 

ACCELERATION CM/SEC**2 



SEABROOK 



Figure 3.2.4b 



BEHCs which include only the contribution to the PGA hazard 
from earthquakes within the indicated magnitude range for the 
Seabrook site, in New England (far from either the New Madrid 
or the Charleston areas). 



-77- 



< 

ui 

>- 

tr 

UJ 



o 

2 
< 



o 

X 

LiJ 



CD 

< 

m 
O 
q: 
o. 



-1 
10 



-2 
10 



-3 

10 



-4 
10 



-5 
10 



-6 
10 



-7 
10 



HAZARD CURVES USING ALL EXPERTS 

I I " ■ I 



I I 

CURVE 1 3.75<MB<5.0 

CURVE 2 5.0<MB<5.75 

CURVE 3 5.75<MB<6.5 

CURVE 4 6.5<MB 




O 

+ 



-* in «o rv 

ACCELERATION CM/SEC* • 2 



WATTS BAR 



Figure 3.2.4c 



BEHCs which include only the contnbut^^^^^^ 
areas). 



-78- 



-1 

10 



10 



HAZARD CURVES USING ALL EXPERTS 






UJ J 

CL 10 



u 

1 

UJ * 

o 10 

X 



t -5 
_j 10 

m 

< 
o 
o 
oc 

o. 

-6 
10 



-7 
10 




•* in *o r^ 

ACCELERATION CM/SEC* '2 

CATAWBA 



Figure 3.2.4d BEHCs which include only the contribution to the PGA hazard 
from earthquakes within the indicated magnitude range for the 
Catawba site near the Charleston area. 



-79- 



45 



40 



35 



30 



25 




in 


o 


in 


o 


in 


o 


en 

1 


1 


00 

1 


00 

1 


1 


1 



Figure 3.2.5 



A plot of the relative location of all the sites in the 
study. The median PGA g-levels with a 10,000 year return 
periods are also plotted. Rock sites are denoted by "R", deep 
soil sites by "D" and shallow soil sites by "S". The relative 
location of the New Madrid (NM) and Charleston (C) earthquakes 
are also shown. 



-80- 




tf, 



6. 



8. 



10. 



IE, 



SITE ID. NUMBER 



Figure 3.2.6 



Plot of the ratio of the probability of exceeding 0.3g PGA for 
the median (line), 85th percentile (plot symbol, "0") and the 
arithmetic mean (plot symbol, "X") for the (shallow soil 
case)/(rock case). Site ID number is the same as the section 
number listed in Table 1.1. 



-81- 



3.3 Regional Spectral Comparisons 

There are three elements which primarily control both the spectral shape and 
level. First, it is the choice of the GM model, here we are referring to the 
major differences between the RV-spectral models and either the Newmark-Hall 
type models or the intensity based models. See Figure 2.4.2. Secondly, the 
local soil conditions have an extremely important impact on both the spectral 
shape and level as discussed in Section 2.2. Finally, the regional seismicity 
can influence both the spectral shape and level. In the preceding sections we 
have examined the influence of factors (1) and (2) mentioned above. In this 
section we want to examine how the regional variation in seismicity influences 
the spectral shape. 

The spectral level is sensitive to both the rate of occurrence and earthquake 
magnitude. The longer period part of the CPUHS is very strongly influenced by 
magnitude. Thus sites which are influenced by very large earthquakes, e.g., 
around the New Madrid region, will have more longer period energy than sites 
in New England where the local activity from smaller earthquakes is 
important. There is some influence of attenuation on the short period end of 
the spectrum, but it is relatively small. 

This is illustrated in Fig. 3.3.1 where we compare the spectral shapes between 
a site where of very large earthquakes dominate the hazard as contrasted to 
a site at which the seismic hazard is governed primarily from smaller nearby 
earthquakes. The Limerick site is a rock site where the hazard is primarily 
from local seismicity and, as discussed in Section 3.2, the hazard at the 
Arkansas site (rock site) is primarily from the larger New Madrid 
earthquake. We see from Fig. 3.3.1 that the main difference in spectral shape 
is at the longer periods. There is some difference at the short period end 
but it is relatively small. 

It is of some interest to compare the spectral shape for the four groups of 
sites in Table 3.2.1. In Fig. 3.3.2 we compare the median 10,000 year return 
period CPUHS between the Seabrook, Arkansas Catawba and Watts Bar sites {all 
rock sites). It is interesting to note that (for the four groupings of sites 
listed in Table 3.2.1), the New England group, the near Charleston group, and 
the half-way between group all have similar spectral shapes. The sites in the 
near New Madrid group have spectral shapes similar to the Arkansas site. 
Figure 3.2.4a, b, c, d shows that the relative contribution of very large 
earthquakes is much greater at the Arkansas site (typical for sites "near" New 
Madrid) than the other sites, thus there is relatively more long period energy 
in the Arkansas spectral shape. 

To some extent, the regional difference between spectral shapes noted in 
Fig. 3.3.1 carries over to the case of soil sites. This is illustrated in 
Fig. 3.3.3 where we compare the CPUHS for the Clinton site (near New Madrid) 
to the CPUHS for the Yankee Rowe site (New England). We see from Fig. 3.3.3 
that there is relatively more long period energy in the Clinton CPUHS than in 
the CPUHS for Yankee Rowe both shallow (i.e. Till-like 2) soil sites. This 
small relative difference is typical. 

-82- 



It is important to note that the regional differences observed in Figs. 3.3.1, 
3.3.2, and 3.3.3 also hold (particularly relative to the median CPUHS) even if 
the low attenuation model selected by G-Expert 5 is not included. If 
reference is made to Fig. 2.3.6, we see that the spectral shape is similar for 
both the case when the low attenuation ground motion is included and the case 
when it is not included. 



-83- 



E.U.S SEISMIC HAZARD CHARACTERIZATION 
LOWER MAGNITUDE OF INTEGRATION IS 5.0 

10000. -YEAR RETURN PERIOD CONSTANT PERCENTILE SPECTRA FOR 

PERCENTILES = 15.. 50. AND 85. 



10 



10 



o 

Ul 
00 






o 



10 



10 



10 



Limerick 



Arkansas 




ltaB^H*MriBAl^>L 



c4 K) "* in lor^ooai 



I o 



r-4 (O ■* irnoi^~axr> 



PERIOD (SEC) 2 



C-l rO Tt IT) lDr-~JXXT> 



Figure 3.3.1 



Comparison of the 10,000 year return period CPUHS between the 
Arkansas and Limerick sites (both rock sites). 



-84- 



LOWER MAGNITUDE OF INTEGRATION IS 5.0 
10000. -YEAR RETURN PERIOD CONSTANT PERCENTILE SPECTRA FOR 



o 

UJ 

I/) 



o 



o 
o 



10 



10 



10 



10 



-1 

10 



CM 

I o 



A - Arkansas 
C - Catawba 
S - Seabrook 
W - Watts Bar 



=^si=^^ 




^^^^^^b^ 



fO ■* in tDh^oocT) 



I o 



K) 't lO CDr^JBcr> 



PERIOD (SEC) 2 



K) -"t lo iDr^/iocri 



Figure 3.3.2 



Comparison of the median 10,000 year return period CPUHS 
between the Seabrook, Catawba, Arkansas, and Watts Bar sites 
(all rock sites). 



-85- 



10000. YEARS RETURN PERIOD 



o 

LJ 






o 
o 



10 



10 



10 



10 



-1 
10 



I O 



I I 

Y - Yankee Rowe 
C - CI Inton 




cj ro -^ mi£>h»aCD> 



c4 to ^ iniorvooji 



CI ro ■* in ijor-^oQcn 



•2 PERIOD (SEC) 2 



Figure 3.3.3 



Comparison of the median 10,000 year return period CPUHS 
between two shallow soil sites (till-like 2). The Clinton 
site (near New Madrid) to the Yankee Rowe site (New England) 



-86- 



10000. YEARS RETURN PERIOD 



o 

Ul 



2 
o 



o 
o 



10 



10 



10 



10 



-1 
10 



CM 

I O 



p ■ 


- Pilgrim 


s - 


- Seabrook 


L - 


- St. Lucie 


T - 


■ Turkey Point 




a K) •* m <or~^ccxr> 



c4 K) ^ ir>u3r-^xxr> 



cj ro ■* ID lOr^oQcri 



I o PERIOD (SEC) 2 



Figure 3.3.^ 



Comparison between the median 10000 year return period CPUHS 
for the sites with the lowest and highest median 10000 year 
CPUHS, 



-87- 



4. SUMMARY OF RESULTS AND CONCLUSIONS 

The results of this study, including Bernreuter et al . (1985), provide the NRG 
with the tools for characterizing the seismicity of the EUS and for describing 
the hazard at any location within that region. These tools are: 

a. A data base of estimates of the seismicity of the EUS and appropriate 
GM models, based on expert opinions', in the form of 

a catalog of maps of zonation of the EUS along with estimates, 
including a measure of uncertainty in the estimates, of the 
seismicity of each zone. 

a catalog of ground motion models including an assessment 

(weights) of their relative merits for propagating the motion at 
the source to motion at any location within the EUS, 

b. A hazard methodology which uses the estimates in the data base (a) to 
develop an estimate of the seismic hazard at any location in the 
EUS. The seismic hazard is described in terms of a hazard curve and 
a uniform hazard spectrum. 

c. A data base of estimates of the seismic hazard computed at the 69 
sites with either operating nuclear power plants or plants seeking a 
license. 

The data base for characterizing the seismicity of the EUS has been developed 
through an elicitation of the opinions of experts in: 

The geotectonic features and seismicity of the EUS. 

Ground motion modeling. 

In using the data base it must be recognized that the results are based on 
information which was available to the experts at the time for the elicitation 
between 1983 and 1987. As additional events occur and more data become 
available these may be a basis for a change in opinion. Thus, it is 
recommended that the NRC consider updating the current data base on a periodic 
basis. This is particularly true of the ground motion models where there is 
considerable activity in development of new and improved ground motion 
models. Methods are also being developed for using the historical records of 
events to complement the opinions of experts about the seismicity. These 
methods have the potential for upgrading the seismicity data base. 

The hazard methodology is based on a probabilistic model of the occurrence and 
distribution of magnitudes of earthquakes and the attenuation of the ground 
motion from a source to a site. It also includes modeling of local site 
effects. The methodology incorporates expert opinion to supplement the 
available data on zonation, seismicity and ground motion modeling. 



-88- 



The description of zonation, seismicity and choice of ground motion model and 
related parameters are assumed to be based on subjective opinions. The method 
assumes that these opinions are expressed in two ways: 

a. A "best estimate" or most likely zonation, model or value of a 
parameter. 

b. A collection of zonations or models with relative levels of 
confidence or a range of values for a parameter which is believed, by 
the expert, to represent the parameter with a high degree of 
confidence. 

From these inputs the methodology can produce two types of estimates of the 
seismic hazard at a site: 

A single "best estimate" hazard curve or spectrum based on the best 
estimate inputs. 

A constant percentile hazard curve or spectrum which represents the 
uncertainty in the hazard as expressed by the 

a. Uncertainty of an individual expert. 

b. Variation in opinions among several experts. 

When using these estimates, it is important that the user recognizes some 
characteristics and limitations of these results: 

1. With respect to the "best estimate" hazard curve (spectrum), it 
should be recognized that this is a weighted average of the hazard 
curves based on the best estimate inputs of several experts. This is 
a single or point estimate of the hazard to which one associates no 
confidence that it represents the true hazard. Thus, it is not 
recommended that this curve (spectrum) be used as an absolute 
description of the seismic hazard at a site. This is particularly 
important since we found considerable variation in the best estimate 
hazards between experts associated with this project. Of course, it 
may be useful to consider point estimate hazard curves for a 
comparison of seismic hazard at different locations. However, even 
then we suggest that the total uncertainty, (due to the uncertainties 
and variations in opinions) in the hazard be recognized in making 
such comparisons. Other point estimates which take uncertainty into 
considerations are also outputs of the methodology. They are: 

The median CPHCs. 

The arithmetic mean hazard curve. 

2. It should be recognized that the CPHCs, which represent the 
uncertainties associated with estimating the hazard based on 

-89- 



subjective opinions are not explicit hazard curves that are derived 
from a unique set of inputs. Neither are they a set of bounds to 
which can be associated a stated confidence, e.g. 90% confidence, 
that the true hazard curve is contained within the bounds. Rather, 
they are locus of the 15th and 85th percentile points, of the 
uncertainty distribution in the hazard, i.e. P(A>a), for each level 
a. Thus, an uncertainty statement, in the sense of uncertainty in 
the inputs, is applicable to the probability of the maximum PGA 
exceeding a specific level a. The same statement is not applicable 
over the entire range of a simultaneously . 

3. Another issue which affects both the point estimate of the hazard and 
the description in the uncertainty in the hazard is the choice of 
weights for combining estimates over the collection of expert 
opinions. Our approach was to calculate the hazard (or the 
uncertainty distribution of the hazard) based on the inputs from each 
pair of seismicity and ground motion experts and then use self 
ratings as a basis for weights for combining the estimates over all 
possible pairs of experts. Appropriate methods for combining 
information from multiple sources, particularly information based on 
subjective opinions, is a complex subject. Several methods, based on 
rankings of experts or strictly mathematical weights, are available 
in the literature. Based on some limitations on the methods we could 
use and the ranking information we could elicit, we chose self 
ranking as the basis for the weights. However, we consider the issue 
of combining information, both from opinions as well as from both 
data based information and opinions as a topic requiring further 
investigation, particularly in light of the sensitivities of the 
results to certain combination of expert's input discussed in 
Section 2.3. 

The detailed conclusions reached in the course of this study are discussed in 
the appropriate sections of the various volumes which comprise this report. 
The following is a summary of the most important ones: 

(1) There is substantial uncertainty in the estimated hazard. The 

typical range in the value of the probability of exceedance between 
the 15th and 85th percentile curves for the PGA is on the order of 40 
times, for low PGA; it is more than 100 at high PGA values. This 
translates into an approximate factor of 4 in ground motion for the 
15th-85th range of values in the PGA given a fixed return period, and 
similarly an approximate factor of 4 in the ground motion for the 
range of values in the PSRV for a given return period. 

The range between the 15th and the 85th percentile hazard curves 
represents the total uncertainty in estimating the seismic hazard at 
a site due to two sources of uncertainty: 

The uncertainty of each expert in the zonation, models and 
values of the parameters of the analyses 

-90- 



The variation in the hazard estimates due to the diversity of 
opinions between experts. 

The latter, or inter-expert variation is an important contributor to 
the total uncertainty in the estimated hazard. Specifically, the 
magnitude of uncertainty introduced by the diversity of opinions 
between experts is of the same order, on the average, as the 
uncertainty in the hazard due to the uncertainty of an individual 
expert in the value of the parameters. However, as shown in Section 
2.4 at times the uncertainty between experts can be very large. 

For a given acceleration value, the range of the median hazard values 
at all the sites analyzed falls within the 15th-85th percentile range 
of any one of those sites. 

(2) The 50th percentile CPHC appears to be a stable estimator of the 
seismic hazard at the site. That is, it is the least sensitive to 
changes in the parameters, when compared to other estimators 
considered in this study. 

(3) The process of estimating the seismic hazard in the EUS is reasonably 
stable. Comparison with our previous results indicated that there 
has not been a major shift in results over the past few years, 
although there have been some significant perturbations in the form 
of recent occurrences of EUS earthquakes and the completion of 
several major studies of the seismotectonics of the EUS. In the 
feedback performed in this study, there were some changes introduced 
by members of both the Seismicity and GM Panels. However, the 
computed hazard when aggregated over all experts did not 
significantly change. However, the introduction of the "new" random 
vibration models introduced a significant change in the spectral 
shape by raising the spectral values in the high frequency range and 
lowering it in the low frequency range. 

(4) It is difficult to rank the uncertainties, because zonation and the 
parameters of the recurrence models are hard to separate. 
Nevertheless, our results indicate that the uncertainty in zonation, 
and ground motion models are more significant than the uncertainty 
associated with the seismicity parameters. The largest contribution 
to modeling uncertainty comes from the uncertainty of the ground 
motion. The correction for local site effects is a significant 
contribution to the overall uncertainty introduced by the ground 
motion models. However, as already noted, the uncertainty introduced 
by zonation and recurrence models is also significant and of the same 
order. 

(5) Based on comparisons between the results of our broad generic study 
and site specific studies, we concluded in Bernreuter et al. (1985) 
that the scale of our study is adequate. No major differences in 
zonation or results occurred between our study and site specific 
studies. 



-91- 



(6) We found, consistent with the conclusions in Bernreuter et al. 
(1987), that generally earthquakes in the magnitude range 3.75 to 5 
would significantly increase the estimated seismic hazard if they 
were included in the analysis. Thus, it may be important to keep in 
mind that the CPHCs and CPUHS presented in this study only include 
the contribution from earthquakes with magnitudes of 5 and greater 
when assessing the seismic safety of brittle components of nuclear 
power plant systems, e.g., such as relays. In addition, in light of 
the discussion given in Section 2.4, it must be kept in mind that the 
PGA value is not a good estimator of the loading that very stiff 
components will experience in the EUS. The actual ground motion will 
be amplified. 

(7) We found that the correction for the site's soil category had an 
important effect on the estimated hazard. In Section 2.2 of Vol. YI 
we provided approximate correction to be applied to the estimated 
hazard for rock site to estimate the hazard for shallow soil 
conditions at the same site. This is useful for sites which have a 
few structures founded on shallow soil. Later in Vol. VIII we will 
provide calculations for the sites with multiple soil conditions. 

Finally, it is difficult to assess if our results have either a conservative 
or unconservative bias. We insisted that our panel members not introduce such 
biases in their inputs and we spent considerable effort in developing a 
methodology which would allow the experts to properly express their 
uncertainty without having to introduce some conservative approximations. 
This was particularly true in the area of regional ground motion modeling and 
in the incorporation of multiple alternatives to account for any local site 
amplification of the ground motion. 

(8) We found that in general the site soil correction is not a linear 
operation on the hazard curve. Thus it is, in general, incorrect to 
modify a hazard curve calculated for a rock site by multiplying by a 
constant number (i.e., mean or median correction factor) to obtain 
the hazard curve at the same site for a different soil condition. 
Performing this incorrect operation could lead to errors in the 
estimate of the PGA, for a fixed return period, by as much as 10 
percent. However, we found that for some sites, multiplying the 
median hazard curve for rock by the median correction factor would 
have given approximately the same median hazard curve we obtained by 
performing the full analysis with our probabilistic correction 
factors. Unfortunately, at the present time, we have not been above 
to develop criteria to identify when performing such operation is 
correct. 

(9) Although the soil site correction is not region dependent, we found 
that other complex interactions, with zonation seismicity and ground 
motion models, made the site correction actually region dependent. 



-92- 



(10) We found that the input from some experts lead to either high or low 
estimates of the hazard at most sites. In particular G-Expert 5's 
input lead to results, in general, higher than when only the other 4 
GM Experts' input is used. 

We found that the impact from any S-Expert did not show a consistent 
deviation from the results of all the other S-Experts at all sites, 
however, the results from some of the S-Experts were found to be 
either high in some region of the EUS (i.e. S-Expert 2), or low (i.e. 
Expert 12, especially in the South West and Central U.S.). 



-93- 



Appendix A 

References 

D.L. Bernreuter, J.B. Savy, R.W. Mensing, J.C. Chen, B.C. Davis, Seismic 
Hazard Chara cterization of the Eastern United States, Vol. 1 and Vol. 2 , 
LLf^L uClD-^0421, Vol. 1 and Vol 2. (April 1985). 

D.L. Bernreuter, J.B. Savy and R.W. Mensing, Seismic Hazard of the Eastern 
United States: Comparative Evaluation of the LLNL and EPRI Studies , USNRC 
Report NUftE6/Cft-4SS§ (1567). 

Lee, V.W. and M.D. Trifunac (1985), Attenuation of Modified Mercalli 
Intensity for Small Epicentral Distances in California , University of 
Southern California Report CE85-01. 

Newmark, N.M. and Hall, W.J., Development of Criteria for Seismic Review 
of Selected Nuclear Power Plants, Nuclear Regulatory Commission Report 
NUREG/CR-0098, May 1978, 49 p. 

Trifunac, M.D. (1986), A Note of the Range o^ Peak Amplitudes of Recorded 
Accelerations, Velocities and Displacements with Respect to the Modified 
Mercalli Intensity , Earthquake Noies 47, pp. 9-24. 



A-1 



Appendix B 

Maps of the Seismic Zonation for 
Each of the 11 S-Experts 



B-1 




I. 

Q. 
X 
UJ 

&. 
o 

40 






o 

N 



E 



3 



B-2 




B-3 




B-4 




ro 



a. 



o 



« 
E 



« 
c 
o 



E 

•r- 
0) 



CD 
0) 



B-5 




i 
c 
c 
> 

u 

1 
c 
<»- 

c 
« 
E 

a 

V 

K 
JC 

C 

o 



c 
c 
o 



E 

V) 

0) 



«l 

O 



B-6 




u> 



a. 



o 



E 

a> 
«/) 



c 
o 



E 



CO 

u 



B-7 




B-8 




o 




o. 
E 
0) 



4-> 



E 
«/) 

•r- 
« 



CD 

0) 

I. 



B-9 




B-10 




B-n 




B-12 




B-13 




B-14 




IM 



U) 



I. 

CL 
X 



O 



a. 
<o 

E 

0) 

«/) 

<e 

c 
o 

•r— 
•4-> 

« 

c 
o 



E 

0) 
t>0 



u> 



Nl 
O 






B-15 




CNJ 



4J 

Cl 
X 



o 



E 






rsj 
o 



CO 



E 




CVJ 
00 

I. 



<\J 



B-16 




rsi 
o 



o 



O 



C^ 



/ 



i/> 



>u>> 



CO 



CO' 



o 



Nl 
O 



ro 



4-> 

a. 

X 

UJ 

t- 
o 



a. 
E 

0) 
«/> 

«e 



o 

N 



E 



ro 

t— < 

CO 

i. 



B-17 




B-18 



NdC rOHM U« U t NUCLIAN Xf OULATOnv COMMIUION 

'*i^^"ijoV BIBLIOGRAPHIC DATA SHEET 

SCf INSTnuCTlONSON THE REVEnSi 


1 RtPQRT NUM8CR IAu'$nfa it TIOC tod Vol Nm.ilntfi 

NUREG/CR-5250 
UCID-2I517 
Vol. 6 


1 TITLE ANOSUITlTLE 

Seismic Hazard Characterization of 69 Nuclear Plant Sites 
East of the Rocky Mountains 

Regional Comparison Between Sites, Site Effects, 

General Discussion, and Conclusions 


1 LEAVE ILANK 


4 DATE REPORT COMPLETED 


MONTH I YEAR 

November ' 1988 


5 AUTHO«(SI 

D.L. Bernreuter, J.B. Savy , R.W. Mensing, J.C. Chen 


6 DATE REPORT ISSUED 


MONTH YEAR 

January 1989 


7 PERFORMING 0R0ANI2ATI0N NAME AND MAILING ADDRESS llociua* Z.p Cootl 

Lawrence Livermore National Laboratory 
P.O. Box 808, L-197 
Livermore, California 94550 


8 PROJECT/TASK/WORK UNIT NUMBER 


9 FIN OR GRANT NUMBER 

A04A8 


10 SPONSORING ORGANIZATION NAME AND MAILING ADDRESS llnc(udtZio Cadml 

Division of Engineering and System Technology 
Office of Nuclear Reactor Regulation 
U.S. Nuclear Regulatory Commission 
Washington, DC 20555 


lU TYPE OF REPORT 

Technical 


b PERIOD COVERED »nc/u«ii«a«»t< 

October 1986-October 1988 


12 SUPPLEMENTARY NOTES "" 


The EUS Seismic Hazard Characterization Project (SHC) is the outgrowth of an earlier study 
performed as part of the U.S. Nuclear Regulatory Commission's (NRC) Systematic Evaluation 
Program (SEP). The objectives of the SHC were: (1) to develop a seismic hazard characteriz- 
ation methodology for the region east of the Rocky Mountains (EUS) , and (2) the application 
of the methodology to 69 site locations, some of them with several local soil conditions. 
The method developed uses expert opinions to obtain the input to the analyses. An important 
aspect of the elicitation of the expert opinion process was the holding of two feedback 
meetings with all the experts in order to finalize the methodology and the input data 
bases. The hazard estimates are reported in terms of peak ground acceleration (PGA) and 5% 
damping velocity response spectra (PSV) . 

A total of eight volumes make up this report which contains a thorough description of the 
methodology, the expert opinion's elicitation process, the input data base as well as a 
discussion, comparison and summary volume (Volume VI). 

Consistent with previous analyses, this study finds that there are large uncertainties 
associated with the estimates of seismic hazard in the EUS, and it identifies the ground 
motion modeling as the prime contributor to those uncertainties. 

The data bases and software are made available to the NRC and to the public uses through 
the National Energy Software Center (Argonne, Illinois). 


Seismic hazard. Eastern U.S., ground motion 

b. lOENTlFIERS/OPENENOEO TERMS 


1S AVAILABILITY 
STATEMENT 

Unlimited 


16 SECURITY CLASSIFICATION 


Unclassified 


{Thtt rtportt 

Unclassified 


17 NUMBER OF PAGES 


18 PRICE 


oU.S.COVERNfflENT PRINTING OFFICE :1999-241.590 .80330 







UNITED STATES 
NUCLEAR REGULATORY COMMISSION 

WASHINGTON, D.C. 20555 



SPECIAL FOURTH-CLASS RATE 

POSTAGE & FEES PAID 

USNRC 

PERMIT No. G-67 



OFFICIAL BUSINESS 
PENALTY FOR PRIVATE USE, $300 



UNIVERSITY OF ILLINOIS-URBANA 





3 0112 084234027